KR20140049087A - Internet preference learning facility - Google Patents

Abstract

In an embodiment of the present invention, an improved ability to target a website response returned to a user through the use of computer equipment by using the user preference learning equipment to determine the user's preference is described.

Description

Computer-readable storage medium {INTERNET PREFERENCE LEARNING FACILITY}

Cross reference of related application

This application discloses the following applications, each of which is hereby incorporated by reference in its entirety, US Provisional Application No. 61 / 233,326, filed Aug. 12, 2009, and US Provisional Application No. 61 / 300,511, filed February 2, 2010. And US patent application Ser. No. 12 / 483,768, filed June 12, 2009.

This application also discloses the following patent applications, each of which is incorporated herein by reference in its entirety, US Patent Application No. 12 / 483,768, filed June 12, 2009, and US Patent Application, filed July 15, 2009. No. 12 / 503,263 and US patent application Ser. No. 12 / 503,334, filed Jul. 15, 2009.

Field

The present invention relates to a collective knowledge system, and more particularly to providing natural language computer-based topical advice based on machine learning through user interaction.

Online searches for topic advice represent a significant use of computer resources, such as those available over the Internet. Computer users currently have access to a variety of search tools to search for advice on specific topics, but doing so may require expertise in the use of a search engine, and may take time to screen, interpret, compare, and compare. You can generate the vast search results you need. People may be accustomed to seeking advice from others in oral natural language, so it may be useful to have a computer-based advice tool that more closely mimics how people interact with each other. In addition, the advice for a topic can change over time, and any static database of advice can quickly become outdated. Thus, there is a need for improved topic advice retrieval capabilities that are adapted for use with natural language and provide continuous content refinement.

The invention may consist of computing equipment, such as machine learning equipment, that can inquire user questions, and based on the user's answers, the system may provide decisions such as recommendations, diagnostics, conclusions, advice, and the like. Internally, the system can use machine learning to optimize which questions should be queried and which decisions should be made at the end of the process. The system can learn through the user providing feedback on the provided decision, including a decision about whether the decision was helpful. Helpful solutions can be enhanced and related to questions and answers queried in this way. When the user indicates that the decision was helpful, the system can remember which questions were asked, what the answer to each question was, and associate these questions and answers with the final decision. In an embodiment, these relationships may be the basis of machine learning to learn over time which questions are asked the next time a user comes to the system.

In an embodiment, the present invention can help a user make a decision through the use of machine learning equipment. The process may begin with an initial query received by the machine learning equipment from the user. The initial question may be received via a search interface, where the final decision is based on the initial search item, the conversation of questions and answers with the user, the training of the system, and the like. The user may then have a conversation consisting of questions from the machine learning equipment and the answers provided by the user. Machine learning equipment may then belong to initial questions such as recommendations, diagnoses, conclusions, advice, and the like, and provide a decision to the user based on the conversation. In an embodiment, future questions and decisions provided by the machine learning equipment may be enhanced through feedback provided by the user.

In an embodiment, the initial question asked by the user may be an objective question, a subjective question, or the like. The questions may be provided from a wide range of topics such as topics belonging to product, personal information, personal health, economic health, business, politics, education, entertainment, environment, and the like. The question may be in the form of a multiple choice question, a yes-no question, a rating, a selection of images, a private question, or the like. The question may be about the user, may be provided by another user, may be provided by an expert, and so on. The question may be based on a previous answer, such as, for example, from a current conversation with the user, from a previously saved conversation with the user, or from a previously saved conversation with another user. The question may be a pseudo random question, such as a test question, a search question that helps select a pseudo random decision at an opportunity where the pseudo random decision turns out to be useful. The question may include at least one image as part of the question. The question can follow the psychological dimension. In embodiments, the question may not directly contact the user, but rather the IP address, the user's location, the weather at the user's location, domain name, relevance to route information, relevance to recent downloads, relevance to recent network access, recent It can be determined from conceptual information, such as through relevance to file access.

In an embodiment, the conversation may continue until the machine learning equipment develops high confidence in the set of reduced decisions presented to the user, such as a single decision presented to the user. The decision provided by the machine learning equipment may be independent of the order of the questions in the conversation. The decision may provide an alternative decision when at least one question of the conversation is omitted, where the alternative decision may be different based on machine learning equipment with less information from the user. The decision may indicate a ranking of decision choices, such as ranking decisions across non-traditional feature dimensions. The decision can display at least one image related to the decision. The decision may be a pseudorandom decision on an opportunity for which the pseudorandom decision turns out to be useful, such as the pseudorandom decision being part of a system of search, where the system of search may improve the effectiveness of the system, and the machine learning equipment may Can be learned from.

In an embodiment, the feedback provided may be related to or derived from how the user answers questions in the conversation, how the user responds to the decisions provided by the machine learning equipment, and the like. In an embodiment, the feedback may be solicited from the user.

In an embodiment, the user can extend the learning of the machine learning equipment by entering new information, where the new information can be his own topics, questions, answers, decisions, and the like. The machine learning equipment can use the new information to determine whether the new information is helpful to the user.

In an embodiment, the expert user may extend the learning of the computing equipment by entering new information, where the new information may be their own topics, questions, answers, decisions, and the like. Expert users entering new questions, answers, or decisions can specify how the system should rank decisions when future users provide specific answers to questions. Expert users can also specify preconditions / dependencies when new questions should be asked. Expert users can optionally also input the importance for the question.

In an embodiment, the system may be implemented as a series of dynamic decision trees. During a question and answer conversation with a user, the system can find out which question is relevant for a query if the user is provided with an answer provided in advance. When a user enters a new question, answer, or decision, they can specify which question relates to which decision and how the answer affects the relative ranking of various decisions.

In an embodiment, the system may include a number of independent, such as how well the decision matches the objective requirements specified by the user during the question and answer conversation, and how well the decision matches the subjective requirements of the user, such as the user's preference. Decisions can be recommended based on the factor.

In embodiments, the user interface may be provided for user interaction with machine learning equipment such as those associated with web interfaces, instant messaging, voice interfaces, mobile phones, SMS, and the like.

In an embodiment, the present invention can help a user make a decision through the use of machine learning equipment. The process may begin with an initial question received by the machine learning equipment from the user, where the initial question may relate to one of a broad category of topics such as product, individual, health, business, politics, education, entertainment, environment, and the like. . The user may then have a conversation consisting of questions from the machine learning equipment and the answers provided by the user. Machine learning equipment may then belong to initial questions such as recommendations, diagnoses, conclusions, advice, and the like, and provide a decision to the user based on the conversation. In an embodiment, future questions and decisions provided by the machine learning equipment may be enhanced through feedback provided by the user.

In an embodiment, the present invention may help a user make a decision through the use of computing equipment. The process may begin with an initial query received by the computing device from the user. The user may then have a conversation consisting of questions from the computing equipment and the answers provided by the user. The computing equipment may then provide a decision to the user based on the feedback collected from the plurality of users. In an embodiment, the computing equipment may enhance future questions and decisions provided by the computing equipment based on the feedback received from the user.

In an embodiment, the present invention can help a user make a decision through the use of machine learning equipment. The process may begin with an initial query received by the machine learning equipment from the user. The user may then have a conversation consisting of questions from the machine learning equipment and the answers provided by the user, where the number of questions and answers provided through the conversation may determine the quality of the decision. Machine learning equipment may then belong to initial questions such as recommendations, diagnoses, conclusions, advice, and the like, and provide a decision to the user based on the conversation. In an embodiment, future questions and decisions provided by the machine learning equipment may be enhanced through feedback provided by the user. In an embodiment, the quality may be high when the number of questions and answers is large, such as more than 10 questions, more than 15 questions, more than 10 questions, and the like. In an embodiment, the quality may be good quality when the number of questions and answers is small, such as less than 10 questions, less than 5 questions, less than 3 questions, one question, and the like.

In embodiments, the present invention may make decisions through the use of machine learning equipment. The system may include machine learning equipment capable of receiving an initial question from a user, conversation equipment in the machine learning equipment providing a question to a user and accepting an answer from the user, machine learning equipment to provide a decision to the user, and the like. In an embodiment, the decision provided to the user may be based on an exchange between the user and the machine learning equipment and may belong to an initial question. The machine learning equipment can also receive feedback from the user to improve future questions and decisions provided by the machine learning equipment.

These and other systems, methods, objects, features, and advantages of the present invention will be apparent to those skilled in the art from the following detailed description of the preferred embodiments and the drawings. All documents mentioned herein are incorporated herein by reference in their entirety.

The following detailed description of the invention and its specific embodiments can be understood with reference to the following features.

1 illustrates a list of topics within a system from which a user can obtain a decision.
2 illustrates example questions that a system may ask a user.
3 illustrates an example video question that a system may ask a user.
4 illustrates an example of the type of information that the system can present to a user when making a particular decision.
5 illustrates an example of a higher list for a camera.
6 shows a second example of a higher list for a camera.
7 illustrates an embodiment of a user home page.
8 and 8A illustrate an embodiment of a user's stored answer.
9 illustrates the choices a user may contribute to expertise.
10 shows an example of a user question.
11 and 11A illustrate an embodiment of an answer format.
12 illustrates an exemplary list of all decisions of a topic.
Figure 13 illustrates an embodiment process flow for the present invention.
Figure 14 illustrates an embodiment process flow for the present invention.
Figure 15 illustrates an embodiment of a block diagram of the present invention.
16 illustrates a contributor / expert interface homepage embodiment.
FIG. 17 illustrates an objective question embodiment for a user seeking help in making a decision.
18 illustrates an embodiment of a decision result indicating a particular recommended decision.
19 illustrates an interface embodiment for a user to establish a relationship between an attribute and a decision result.
20 illustrates an embodiment illustrating how a user can edit the decision result.
FIG. 21 illustrates an embodiment illustrating a change between a previous modification of content and two previous modifications of content. FIG.
FIG. 22 illustrates an embodiment representing a question edited by a user. FIG.
Fig. 23 is a diagram showing an embodiment showing a modification history for an attribute.
FIG. 24 illustrates an embodiment of a workshop interface in which areas of newly added advice may be displayed.
25 illustrates an embodiment in which the system asks the subjective question to the user to learn the user's preferences.
FIG. 26 illustrates an embodiment representing an activity feed of recent activity by a contributor.
FIG. 27 illustrates an embodiment representing results based on multiple dimensions. FIG.
FIG. 28 illustrates an embodiment representing multiple question and answer results in response to a user's unstructured input.
FIG. 29 illustrates an embodiment representing an example question to a user inquiring about their preferences. FIG.
30 shows a similarity profile of news celebrity Glenn Beck as determined in an embodiment of the invention.
FIG. 31 illustrates a similarity profile of celebrity Martha Steward as determined in an embodiment of the invention. FIG.
FIG. 32 illustrates an embodiment for using a third party API to help learn about a user. FIG.
FIG. 33 illustrates an embodiment for using third party APIs to help learn about a user and to target responses returned from the user query to the user.
FIG. 34 illustrates an embodiment for determining preferences of an unknown user through the use of an internet social interaction graphical representation.
FIG. 35 illustrates an embodiment for improving user preference and preference profiling. FIG.

Although the present invention has been described in connection with certain preferred embodiments, other embodiments may be understood and included by those skilled in the art.

All documents referred to herein are hereby incorporated by reference.

The present invention may query the user 1314 with a question 1320, and based on the user's response, the system may provide a decision, such as a recommendation, diagnosis, conclusion, advice, or the like. Internally, the system can use machine learning to optimize which questions 1320 should be queried and which decisions 1310 should be made at the end of the process. The system may learn through the user providing feedback on the final decision as to whether the decision 1310 was helpful. Helpful solutions can be enhanced and related to the questions 1320 and answers 1322 in this way. When the user 1314 says that the decision 1310 was helpful, the system can remember which question 1320 was asked, what the answer 1322 for each question 1320 was, and these questions 1320 ) And the answer 1322 may be related to the final decision. These relationships may be the basis of machine learning to learn over time which questions 1320 are asked the next time a user 1314 comes to the system.

For example, user 1314 may attempt to get advice selecting what to visit. The system can ask the question "How old are you?" And get an answer "30s." Finally, the system may display the decision "Kelley's Irish Bar" to the user 1314. Assume that user 1314 says that this decision was helpful. The system will increase the association between the question "How old?", The answer "30s" and the decision "Kelly's Irish Bar". The next time the user 1314 comes to the site in search of advice on the bar, the system asks the user 1314 a "How old" question (1320) because this question (1320) was useful to help the user in the past. You will be more likely to contact us. If the user 1314 answers the question 1320 in the same way as the previous user 1314 ("30s"), the system will increase its confidence that the final decision is "Kelly's Iris Pub".

The system can build a profile of the user's preferences, aesthetic preferences, and the like, and learn via feedback which decision 1310 is liked by which type of people. Alternatively, the expert user may specify which kind of decision 1310 will be liked by what kind of person. Learning the user's preference profile may occur through separate processes from conversations of questions 1320 and answers 1322 inquired by the system at a particular topic. For example, user 1314 may speak to the system individually about their preference selection through different question and answer conversations specifically designed to understand the user's aesthetic preferences.

The user 1314 may not want to spend time to educate the system about all of these preferences, so instead the system will ask all to consult in the context of the user 1314 making one particular decision 1310. The preference or question 1320 can learn or designate an expert which is the most important preference or question. From the world of all questions, the system can be aware of what it finds about the palatability profile, for example the system has three specific questions 1310 that are best when the user 1314 attempts to find a sedan of less than $ 25,000. You may be learning that this exists. Alternatively, there may be an entirely different set of three preference questions for a user 1314 interested in an SUV of more than $ 45,000.

User 1314 may also speak to the system only for their preference preferences, not for any objective question. In this case, the system can provide a ranking of all decisions 1310 in the area of advice based purely on preference. Thus, instead of the user 1314 saying that they want a $ 200 point-and-shoot camera, what the user 1314 would effectively do is a camera that another 35-year-old urban man would prefer a computer over sports. Say you want it. The user 1314 can dictate this preference by using the search interface and selecting an area of advice that is explicitly labeled as "Camera for 30 Urban Men" instead of the "Which Camera to Buy" advice area. Alternatively, the user 1314 may direct their interest in making a decision 1310 for the camera, and then choose not to answer any questions in the Q & A conversation from the system, so the system It may have only subjective information about the user 1314 for use in recommending a camera to the user 1314. Alternatively, user 1314 may answer questions 1320 in the conversation, both objective and subjective, and the system may then recommend the camera based on the combined objective data for the camera and subjective data for the camera. Can be.

The user can also enter new questions, answers, and final decisions. The system may attempt a new question 1320 for future users to see if the question 1320 turns out to be useful for helping these users. For example, the user 1314 of the bar recommendation service may post the question "Do you want a noisy place or a quiet private setting?" The system may decide to query this question 1320 through the process described above that observes the correlation between the answer of this question 1320 and the recommendation that the user found useful in future use of the bar recommendation service. On the other hand, user 1314 may contribute a question 1320 that has no value in helping the user. For example, user 1314 may contribute the question "Do you have a Canon camera?" The system may try this question 1320 for future users and fail to note any correlation between the recommendation and the answer of this question 1320 as the user finds it helpful. In this case, the question 1320 may be less queried because there is no prediction of whether one recommendation or another was helpful.

The system may continue to ask questions 1320 if it does not feel highly confident of a few possible decisions. The system may also stop soon if it has already asked too many questions 1320 and feels there is a risk of angering the user. The system may also query at least the minimum number of questions 1320 to evade the user 1314 if the system feels that the system may not be able to inquire enough to make intelligent decisions.

The system may have a mechanism to withstand inaccurate answers from the user. Inaccurate answers may result from a user 1314 who does not understand the question, does not understand the answer 1322, or does not recognize the answer 1322 to the question. If most of the answers provided by the user 1314 support a particular decision, then the system can make a decision 1310 even if all the user's answers do not support this decision.

In an embodiment, the present invention may provide at least one of a question 1320 and an answer 1322 between the system and the user, machine learning used to improve the decision and decision to the user. The system includes an improved way to generate questions 1320 and answers 1322, an improved way to provide decisions to users, and an improved way to use machine learning to improve decisions and questions 1320 provided by the system. Etc., any of these capabilities may be used individually or in combination as a standalone system or may be integrated into a third party system as an enhanced capability. In an embodiment, each of these enhanced capabilities may utilize some form of machine learning as described herein. For example, the system may provide an improved way for the user 1314 to execute the questions 1320 and answers 1322 by learning under what circumstances the user 1314 seeks specific information. For example, the weather can be learned by the system that the user is a condition that the user has different preferences depending on the date and time and weather conditions. When it rains during the day and the user 1314 searches for a movie, the user 1314 may be more likely to find a movie ticket and the location where the movie is shown. When it rains at night and the user 1314 searches for a movie, the user 1314 may be more likely to find a description of the movie. In another example, the system learns, for example, that a user prefers a particular format during the day versus during the evening, provides a single decision 1310 versus selection based on age, and determines based on the geographic location of the user ( It is possible to provide an improved way of providing a decision to the user, such as preferring more questions 1320 prior to the presentation of 1310. In another example, the system uses more professional information based on age and education, for example, when the topic is fashion and the user 1314 is younger vs. more popular opinions versus the user 1314 is older. To learn which decision 1310 is chosen for the user, such as using more of the traditional practice at times, asking more questions 1320 about the user's choices among friends when the topic is personal, etc. It can provide an improved way.

In an embodiment, the invention provides a question 1320 for providing a decision and learning which decision is provided if one of the elements is not provided by the system, such as when this element is provided by a third party system; You can provide a combination of answers. For example, a third party search engine web application may be to enhance their ability to provide a categorized list from a user's search query, and thus to query their query 1320 to reinforce their keyword search and classification algorithms. And may use the equipment of the present invention to generate an answer 1322. In this case, the third party search engine provider may not be interested in the equipment of the present invention for generating the decisions because their services are in the business of providing a sorted list rather than a limited set of decisions. However, the present invention may allow the search engine provider to improve this classification result to the user based on the present invention's ability to continuously improve the questions 1320 and answers 1322 to the user. In that sense, they can provide significant new capabilities to search engine providers.

In an embodiment, the subject of the initial area of advice can be specified via a search interface. For example, a user 1314 searching for "Romantic Honeymoon in Italy" would have a user 1314 visit Italy instead of first asking user questions about where they want to go and what type of vacation they are looking for. Can be directed to a web page that helps determine where to go on a honeymoon. Or, the user 1314 can search for a specific location in Italy and 1) help the user 1314 determine whether this particular location is good for their needs (e.g., "This vacation is a honeymooner And "good for a romantic getaway" or "bad for a family vacation"), 2) starting a conversation to help the user 1314 find an alternative potentially better location in Italy for a vacation. You may be directed to a web page that provides. Alternatively, user 1314 may search for a particular product and then enter a conversation to narrow down which of these products is best for them. In both cases 1) and 2), the displayed information may be based on how other users answered the question in the decision making conversation and then provided positive feedback for this decision. Therefore, if many people using the topic "Where to go on vacation" answer "Yes" to the question "Do you want a romantic vacation" and then provide positive feedback in "Italy", the system will It may indicate that Italy is a romantic destination for a user 1314 coming via a search engine. Alternatively, the user 1314 who adds the decision "Italy" or "Do you want a romantic vacation" to the system answers the question "Do you want a romantic vacation" with "Yes" related to Italy And thus Italy may explicitly instruct the user 1314 to come via a search engine to indicate that it is a romantic vacation.

In an embodiment, the present invention uses, for example, the equipment of the invention to make better decisions, as well as using input from experts, and sessions of questions 1320 and answers 1322 for a particular user. Using the previous user 1314 interaction with the system to provide the decision as well as to provide, and to the user 1314 to answer questions 1320 and answers to the system to learn about future decisions. 1322, as well as providing a reward to the user 1314 rather than a decision, making a decision 1310 without any learning, such as querying the questions 1320 and answers and simply filtering the results, how This ability not only leverages the ability of the present invention to learn whether to make better decisions, but also enables the expert system to do more with the user through the question 1320 and answer 1322 interfaces. Other combinations of some subsets of learning to ask questions, make decisions, and make better decisions may be provided, such as to provide a. In an embodiment, the system can provide a user 1314 session for all elements of question 1320 and answer 1322 to provide a decision to the user and learn how to improve the decision.

In an embodiment, the user 1314 entering the question 1320 can optionally specify a dependency and importance on the question. Dependencies can control when questions can be queried. The importance may specify the relative importance between the different questions 1320 to weight the user's 1314 answer. If a compromise is to be made because the system does not match any decision 1310 results to all answers 1322 specified by user 1314, the system will attempt to recommend a decision result that matches the high importance question to the low importance question. Can be. The system may also prioritize asking high importance questions for low importance questions. For example, a user who enters a new question, such as "Where would you like to take a vacation in the United States," would be "world" before answering the new question "Where would you like to take a vacation in the United States"? Set dependencies that require existing questions, such as "Where do you want to go?"

In an embodiment, the present invention may provide a system with a user 1314 interface through which a user 1314 can interact with equipment of the system. The system may include a number of parts, some of which may be a collection of websites, supervisors, and widgets. A widget may be a collection of code that collects, processes, and renders a single piece of content on a website. The website may consist of interfaces for end users, staff members, and registered users to make decisions, edit decisions, and view reports on system performance. The supervisor may be a container for executing the widgets such that the widgets are processed prior to user 1314 requests to perform time consuming data collection and render this content.

For example, the widget may collect video about the decision from the internet. The widget in the supervisor can crawl the web viewing the video for each decision 1310 and store the video he found in the database. When the user 1314 comes to the website and gets a specific decision, the website can request that the video widget render itself and display any video he previously found.

Multiple instances of the supervisor can run on multiple computers to scale widget processing. Each widget can itself run on a computer. Similarly, multiple computers may provide an interface to the system via a web server, instant messaging, voice gateway, email, programmatic API, embedded within a third party website, and the like.

In an embodiment, the attribute may be a combination of the question 1320 and one specific answer 1322 for this question. For example, if the question 1320 is "How old?" And the answers to this question 1320 are "Under 18", "20-30" and "Over 30", how old is the attribute? Younger than 18 years old ". The system can operate by learning the relationship between attributes and decisions. When the system queries the question 1320 and the user 1314 provides the answer 1322, the system can take this attribute and know which decision is related to it.

In an embodiment, the system may understand that some represent continuous values and others represent discrete values. When using continuous attributes, the system is often permissible to recommend products that are less expensive than the user 1314 requires, but it is almost acceptable to supply products that are more expensive than the user 1314 requires. It is possible to make more intelligent trade-offs, such as to understand what is not.

In an embodiment, the relationship between the attribute and the decision may be learned from the user and explicitly provided to the system or some combination of the two. For example, a price attribute of “How much are you willing to spend? Less than $ 200” can be explicitly linked to cameras within this price range based on data from experts, e-commerce sites / APIs, and the like. However, the relationship between the attribute "How to use the camera? On vacation" and possible vacation destinations can be fully learned.

When entering a new question 1320, an answer 1322, and a result, the user 1314 can optionally specify a relationship between the attribute and the decision result. For example, if a user 1314 enters the question "How much do you want to buy?" On the topic "Which camera should I buy?", The user 1314 answers that "less than $ 200" is related to cameras X and Y. It can also be specified to the system that it is not associated with camera Z. Next, future users use the topic "Which camera should I buy" and answer the question "How much are you willing to spend"? Answering "less than $ 200", user 1314 recommends cameras X and Y rather than camera Z. You can have a higher chance of being.

After seeking advice from the system and receiving a decision result, the user 1314 may also be provided a reason from the system as to why this particular decision result was recommended. This description may also allow the user 1314 to change the attribute for the decision result if the user 1314 considers the decision result to be recommended as an error by the system.

In general, a learned relationship may involve learning from an automated data feed or some combination from a user, expert, employee, third party.

In an embodiment, there may be various ways in which the system may recommend a solution and select the next question 1320 to ask the user. Possible machine learning systems can be geometric systems such as nearest neighbors, vector machines, geometric systems related to decision trees, Bayesian inference, random forest, boosting, logistic regression, multi-faceted navigation, query improvement, query expansion, singular value decomposition, etc. It can support evolutionary systems such as stochastic systems, genetic algorithms, decision trees, and neutral networks. These systems include complete gameplay (eg, all attributes provided by user 1314 before making a decision), a subset of the answers to individual questions / gameplay, one positive feedback, only negative feedback or their It can be based on learning from some combination. In addition, the system may have previously interacted with the user 1314, such as remembering a previously answered question, a decision that the user 1314 liked or disliked, and which areas of advice the user 1314 had previously advised. Whether it was pursued or not. In addition, the system may be configured by the user 1314 such as the date and time the user 1314 used the system, the user's IP address, client type (e.g., Firefox, IE, mobile phone, SMS, etc.) and other such data. An implicitly provided factor may be considered.

In an embodiment, the invention is directed to question 1320 instead of implicitly learning based on the user's behavior, which may be even more powerful, for example because the system does not remain attempting to infer the user's intentions, moods, and the like. It is possible to provide a machine learning system that goes well beyond the power of collaborative filtering, such as by explicitly querying. In addition, selecting a question 1320 to contact the user 1314 based on what they have already answered may allow the present invention to focus on nuances that may otherwise be omitted. The present invention may have the ability to describe a decision, such as providing a decision beyond simple extrapolation of past behavior, such as, for example, 'others who also purchase X, Y and Z who also like Product A'. have. Instead, the present invention may allow the user 1314 to say that user 1314 should do A because they say they want X, like Y, and believe in Z. In addition, the present invention allows users to contribute new questions 1320 that may be useful, and then questions 1320 automatically learn under helpful context. In other areas of difference, the machine learning techniques of the present invention may be capable of providing decisions in a wide variety of user 1314 areas of interest, where cooperative filtering has difficulty applying to non-product / media applications. For example, collaborative filtering may require a user 1314 on a very personal topic, such as a rare question 1320 such as whether they need to get a tattoo or whether certain costs can be deducted on a user's tax return. It may not be readily applied to assist in making decision 1310. The present invention may be capable of this application. In an embodiment, the present invention may be able to use pre-programmed expert advice intermixed with advice learned from a group of users to make decisions to the user.

In an embodiment, the system may have a wiki web site for editing all data on the system. The web interface can be used to edit / create / delete questions, answers, attributes and solutions. Each solution may also have various information associated with it that may be displayed on a decision page when the solution is recommended. For example, when recommending a vacation in Cancun, the recommendation page may display a video about Cancun. All this ancillary data about the solution can also be editable through the wiki.

In an embodiment, the wiki can be used to edit the data collected by the widget when running in the supervisor. This allows the widget to collect data in advance and then have a human quality assurance process to review and change the collected data.

In an embodiment, the system may maintain a history of all changes made by the widget or human. For example, one use of this history may be to review the work performed by the hired contractor who makes content quality assurance. Another use of this history may be to ensure that the widget does not undo the work performed by humans. For example, if a widget collects a particular video and a human deletes it because it is inappropriate, the widget can use the history to prevent re-adding this video sometime in the future. Finally, if the data is tampered with or incorrectly deleted, the history can allow a means of retrieval.

In an embodiment, when the widget discovers new content, they may queue the task in a human workflow to validate and edit this content.

In an embodiment, the system may sometimes make a random or anti-random decision in anticipation of recommending something that the system may not be useful but which may prove useful. If the system wants to use what has already been learned, it may not make a random selection of which question 1320 is queried and which decision 1310 is made. There may be a compromise between using what is already known, also called exploitation, and potentially learning something new, also called exploration. Utilization can lead to more satisfied users, while navigation can make the system smarter.

In an embodiment, one way to perform this compromise when selecting a question 1320 to query the user 1314 is useful for the system to make a decision 1310 and then retrieve a few random questions 1320 to be queried. It may be to ask a question 1320 that is certain to be. Another way of making trade-offs may be that a fixed set of questions 1320 is based on utilization and the next set has a fixed budget in all user 1314 interactions based on navigation.

In an embodiment, the determination may also be searched or utilized. If the system wants to learn, it may indicate a random decision. Instead of displaying purely random decisions, the system may also display a decision 1310 that meets some needs specified by the user 1314 and searches purely within the remaining requests. For example, instead of taking out a random camera for display to the user 1314, the system may retrieve a random camera that meets the user's price requirements. This may result in more efficient training because the system may be less likely to indicate a decision that does not have the opportunity to meet the needs of the user. Rather than displaying a random decision 1310 at the time of searching, the system may also display both the utilized decision 1310 and the searched solution and obtain feedback separately from the user, respectively. Alternatively, the system can inject a limited amount of randomness and take a decision such as "what is the best guess of the system." For example, the system may predict whether the user 1314 may like one particular camera, but instead make similar but identical other balances to make the appropriate decision 1310 and still learn new information from the user. You can recommend a camera that doesn't. In an embodiment, the system may or may not identify the user 1314 when querying the question 1320 or making a decision through search versus utilization.

In an embodiment, the system can be observed as examining the user for the various things it recommends. For example, the system may ask the user 10 questions 1320 about the Canon SD1000 camera. This can provide a rich set of data for each camera so that the system can begin to build a list of what kind of users 1314 are likely to like this camera. The system may build a ranked list of decisions for each attribute, for example, from the most likely to be connected, given the properties, to the least likely to be connected. For example, the system can build a list of cameras in the order in which they are likely to be connected by someone who says, "How old? Over 50." This may be indicated by the system as the top 10 cameras for users over 50 years old. Many of these top ten lists may be constructed based on the data of the system. These lists can also be combined to form a new list. For example, given a ranked list of cameras for the attribute "How old? Over 50" and another list for the attribute "Why buy a camera? Travel", the system will ask for a Configure a new ranked list of cameras for " user ". These combinations of parent lists may be pre-generated on demand by incrementally asking the user 1314 to select a new parent list or the like.

In an embodiment, these "parent lists" can be used for various purposes. Some users may not want to answer a series of questions 1320 before receiving a decision. Instead, they may be able to browse through these lists and find relevant decisions. The system may have multiple parent lists, such as thousands or tens of thousands, each of which may have its own web page. In addition, these pages may contain a large amount of content that is indexed by search engines and can direct users to the system's website. Alternatively, user 1314 may use a search interface in the system itself to find the area of advice they want to determine. The various parent lists may be used to shorten the conversation by implicitly answering some of the questions 1320 in the conversation based on the parent list. For example, from the area of advice called "vacation" and the already answered "vacation" conversation, a number of questions (1320), "Where do you want to go? Europe", "Where do you want to go in Europe? Italy There may be a parent list called "Romantic Honeymoon Vacations in Italy" that serves as a shortcut or gateway into the "vacation" topic with "," Do you travel to a specific event? Yes "," What is a specific event? Honeymoon "? have. This may serve as an alternative interface for the user 1314 to seek advice through the traditional search interface without involving the question and answer conversation.

In an embodiment, the various pages on the site may have a built-in display of information called widgets. For example, the decision page may include widgets that show how other people who like this question 1320 answer a variety of questions, videos / pictures of the decision, links to other websites with information about the decision, how users Personalized pros and cons of this decision (1310) based on whether (1314) answers the question, a list of other decisions that may be made that the question (1320) is answered differently, the product / honor of this decision Lists (such as recommended consumption reports).

In an embodiment, the system allows the user to navigate through the field of decision (eg, camera, vacation destination, etc.) along dimensions that are not typically available. For example, instead of displaying a camera and just letting the user 1314 say "show me a higher / lower camera," the system shows that the user 1314 shows a "younger person's favorite camera." "," Show me a better and less fashionable camera for my trip. " Dimensions such as "style", "good for travel", "bad for young people", etc. can be created as side effects by querying user questions 1320 and then learning that good decisions 1310 provide these answers. have.

In an embodiment, navigating along an alternate dimension may be used as a starting point for the user 1314 instead of selecting an area for the user 1314 to seek advice and then engaging in a conversation. The user 1314 may begin interacting with the system by using a search interface or an external search engine to retrieve specific decision results such as product names or travel destinations. The system then displays user information for a particular decision result, allows the user 1314 to adjust other decision results, engages in a conversation to detail what the user 1314 is looking for, or the system Allows display of user 1314 information that learns (via machine learning, expert device, or some combination) about the outcome of the decision. For example, user 1314 may use a search interface to navigate to a web page that displays information about a Canon SD1100 camera. The system can display other cameras that those looking for the Canon SD1100 also like, and allow users 1314 to find similar cameras along non-traditional feature dimensions, such as better cameras for shooting sporting events, In addition, the system is recognized for the Canon SD1100, such as "Good for travel", "Not good for those who are learning photography", "Available below S200", and these are preferred by those considering design consciousness ". It is displayed.

In an embodiment, another possible interface may be to display a list of decisions to the user and to display a brief description of why each decision 1310 is made. For example, when recommending a camera, the system may display three cameras, one "cheap", one "longer zoom", and the other "better to travel." This may help the user 1314 see alternatives that the user may not know based on how they answer questions 1320 to reach decision 1310.

In embodiments, users may be asked different types of questions, such as questions 1320 about recommended items (prices, colors, etc.), questions 1320 about themselves, and the like. The system can differentiate users along dimensions such as psychographic and demographic dimensions. User characteristics that may be predictable include age, gender, marital status, whether they live in rural / urban areas, frequency of church attendance, political affiliation, aesthetic preference, satirical / humor sense, socioeconomic background, preference, It may include preferences for a neat or confusing lifestyle, and how much they plan ahead.

In an embodiment, it may be difficult to query the question 1320 directly, and instead the system may attempt to measure those that are correlated instead. For example, instead of inquiring about income, the system can inquire which store (e.g., Wal-Mart, Target, Sox, etc.) the user 1314 prefers. Aesthetics may be determined by displaying images, living rooms, clothing, and the like, and inquiring which style the user 1314 prefers. In an embodiment, the burn may take the place of the question (the answer may be about how you respond to the burn) or the burn may be a question 1320 such as "which is most similar to the clothing you wear". ) Can be used instead.

In an embodiment, the system can group the questions 1320 by whether they are for the item being recommended or for the user. The system may describe which types of questions 1320 are queried to help the user 1314 inquire of potentially aggressive questions 1320 and other ways to understand the surprising value. The system also asks the user 1314 how many questions 1320 are left while asking the question, and the user 1314 by telling the system in advance which decision 1310 is to be made. Other types of messages may also be displayed to the user 1314 as abusive.

In an embodiment, an instant messenger (IM) system may provide a natural interface to the system's question 1320 and answer 1322 conversations. For example, user 1314 may invite the system of the present invention to their "friend list" and then initiate a conversation to obtain decision 1310 via IM. The system can IM the first question 1320 to the user, and the user 1314 then eventually when the system IMs the user 1314 a link to the decision or directly IMs the name of the decision 1310 to the user. May be able to IM these answers 1322 again. In embodiments, other forms of recommendations may also be used, such as cell phones, SMS, emails, and the like.

In an embodiment, a system, for example in the form of an application, may be embedded in a third party web site. For example, the system may be deployed on a website that sells cameras and offers users to recommend related cameras. Alternatively, after user 1314 searches for cameras and has a list of potential cameras of interest to them, the system may ask questions 1320 to help the user 1314 determine from a list of cameras. Can be. For example, if all the cameras that user 1314 considers are good for travel, the system may not ask how the user 1314 wants to use the camera, but the system may require interchangeable lenses. It can be realized that inquiring whether it can be used to recommend one camera over other cameras.

In embodiments, the system may include products (eg, cameras, TVs, GPS / navigation, home audio, laptops, bathrooms and beauty, babies, gardens / outdoors, cars, jewelry, watches, clothing, shoes, etc.), travel (eg For example, wherever you go, where you stay, where you visit, what you do there, etc., finances (e.g. which mortgages, funding needs, which credit cards, something deductible from taxes) , What types of IRAs are saved, asset allocations to capital, etc., gifts for various holidays and events, other date-based decisions (such as what to dress for Halloween), personality (e.g. About their relationship, their career, etc.), recommendations of the right pet, drinks and other aspects of night life, books, movies, film, music, concerts, TV shows, video games, where to eat, what Ordering, User (1314) Live the most similar and related artists, gift recommendations, such as celebrities, some neighbors, it is possible to determine from the plurality of topic areas, such as in the television sees what.

In an embodiment, the system may include areas of technology / IT (eg, computers, software, printers, home networking, wireless, business networks, performance issues, etc.), medical / health, automotive, relationship or interpersonal issues, home and building issues. It can be used to diagnose problems such as back.

In an embodiment, the user of the system can be an anonymous or logged in user. The logged in user 1314 may be a person who creates an account on the site. The logged in user may also have profile pages for them. The content on the profile page is the basic information about this user (nickname, picture, etc.), the decisions they have received and connected, and the system that the user 1314 would like even if the user 1314 does not answer the question 1320 in the topic area. This predicted decision, a list of facts about the user 1314 that the user 1314 did not provide (eg, the age of the user or so that they do not have to be repeated each time the user 1314 uses the system for the decision) Their aesthetic preferences can be provided once and remembered across different times when the user uses the system), a task that the system considers that the user 1314 may be interested in quantifying and doing via the wiki List of [e.g., new user 1314 reviews submitting content, corrects spelling errors of user 1314 submitted content, and That would like] to review the new content discovered by, and may include others, such as having similar answers to questions.

In an embodiment, a user may also have various titles, ranks, or levels on the system that can affect what they can do. For example, some users may be provided with a title of "mediator" of a particular topic that allows these users to edit certain aspects of these topics. Ranks and titles determine how many decisions they provide, how many new questions (1320) or solutions they contribute to the system, how many tasks they accomplish using the wiki, and how good are their specific questions (from their various topics) 1320 may be assigned manually or through automation means including based on answering 1320 or the like.

In an embodiment, a user who is not logged in may not have the benefit of using a system with a large selection of aesthetic or preference based preferences already entered into their profile. Based on learning or manual training from the logged in user 1314, the system may select some aesthetic questions to ask in the conversation when a user who is not logged in seeks advice in a particular topic area. For example, based on a logged-in user who answers preference questions about himself and then provides feedback about which cars they like and dislike, the system asks 1320 whether they enjoy gourmet meals. You can learn how useful it is to contact non-login users who are trying to make a decision between Toyota and Lexus. Using attribute associations learned or manually specified by the logged-in user, the system can then adjust whether to recommend Toyota or Lexus to the non-logged-in user.

In an embodiment, the system can learn from the user who submits feedback in the decision. Some users may intentionally or unintentionally provide inaccurate feedback. For example, a seller may attempt to play a game on the system to get their products highly recommended. Alternatively, a user 1314 who does not know more about the video game may recommend a video game that is not actually a good video game. The system may attempt to filter out feedback from these users by various means. The system may throttle the number of feedback that a given user 1314 can submit (and have a higher throttle limit if the user 1314 logs in or has a high rank / title). The system also shows how well the user (" test " question 1320 during the question 1320 and answer 1322 is tested to test the user's knowledge of subjects and weighted feedback from users with more insight than users without insight. 1314 may throttle or weight the feedback based on whether it answers. The system may also require the user 1314 to pass through a 'captcha' (Completely Automated Public Turing test to tell Computers and Humans Apart) before their feedback is counted or they get a decision. The system can also find the series of answers provided by the user 1314 and weight the user's feedback based on the series of answers. For example, if user 1314 always clicks first answer 1322 or user 1314 clicks in a very unlikely manner, the system may weight the user's feedback low. Finally, the system may determine not to change the weight of the user's feedback or to display decision 1310 based on the history of previous game play. For example, at the tenth time a user 1314 attempts to obtain a camera decision 1310, the system may weight their feedback less than the ninth.

In an embodiment, the system may include search engine optimization (SEO), a process that improves the system's website ranking within the main search engine. This process involves finding keywords that a user is searching for, understanding the competition of search engines to have pages of the site that users come to when searching for these words, understanding how search engines rank sites, Can be categorized into a number of nearly automated steps, such as understanding whether changes to the website need to be done to increase the ranking of the site for a common search.

In an embodiment, finding a keyword that a user can search may be performed using data for historical searches licensed from third party data providers, using, for example, keyword suggestion tools such as those provided by Google and Yahoo. It can be found through different means, such as crawling other websites to know the words they use. Once these keywords are found, the system bids on these words via search engine marketing (SEM), expects to get search traffic in the future, develops content on the system's site for these keywords, and how We can use the data in a number of ways, such as finding if our competitors use these same keywords.

In an embodiment, the system may understand how other sites do and how they rank in the search engine by running keywords through a search engine and finding who is advertising each keyword and which top natural search results are for each keyword. have. Sites discovered through this process can be crawled to discover more potential keywords. The system may also decide to develop new content or evade the market based on this competitive information. If there are a few high ranked sites in the content area, the system can develop content in this area.

In an embodiment, the system may understand that the paid advertisement that leads the user 1314 to the site is relatively inexpensive in one topic area of advice on the site and expensive elsewhere. Thus, the system advertises for low cost traffic, helps their users 1314 with their decisions, and then recommends that these users 1314 use the system in expensive topic areas to advertise and purchase traffic. You can try. For example, the system can determine which dog breeds they should buy, help these users 13114 determine which dog breeds are best for them, and then find out where they should buy their pet medications. You can run ads to people who want to guide you. The latter topic area may be expensive for the system to source traffic due to expensive advertising charges, while the former topic area may be competitive for customers who want to advise what type of dog some existing business will get. It can be relatively inexpensive because it can be.

In an embodiment, the system can understand how search engines rank their natural (unsponsored) search results by studying the relationship between the sites that come when a search is made and the factors of those sites. Possible factors that can be correlated between sites that come with a high ranking may be factors such as the content of the site, the number and quality of other sites connecting to the site, the type of content on these other connected sites, and the like. From the previous step, the system can generate a list of site factors that are ranked by their ability to increase the site ranking of the search engine, and the like. The system may then use this ranked list to make changes to the site to increase the likelihood that the site as a whole or a particular page on the site may be ranked higher in the search engine.

Search engines typically use the keyword index to find documents that are related to a user's query. In an embodiment, the present invention may utilize a "decision index" that may also map user input to related literature. The index can be built automatically, the expert can build the index manually, and the index can be learned through feedback from different types of users who implicitly or explicitly decide to train the system. The results of the search using the decision index can be displayed as a list of documents, a single document, and the like.

Referring to FIG. 1, a user includes a camera, a mobile phone, a coffee and an espresso, a drink, a favorite celebrity, a GPS device, a grill, a Halloween, a laptop, a personality, a toework, a TV, a vacation, a video game, a watch, and the like. An embodiment is provided for a list of topics 102 in the system that can be obtained. In addition, there may be an indicator of the number of decisions learned from user ratings as learned from 43,921 user ratings (104).

Referring to FIG. 2, an embodiment of an example question 1320 is provided that allows the system to query a user 1314. In this example, user 1314 requests a decision regarding the purchase of a camera, and question 1320 is, "How much are you willing to spend?" The user 1314 can now select from the selection 204, such as choosing between less than $ 200, up to $ 300, up to $ 500, over $ 500, unknown, and the like. In addition, there may be an indication as to how many questions 1320 can be queried, such as "in ten or less questions, obtaining preferred camera decisions by a person like you" (202). In embodiments, the user may also provide their own questions, their own answers, their own decisions, and the like, where the system may use this information in current or future decision sessions. In an embodiment, the user 1314 can choose to skip the question 208, where the user 1314 can now be provided an alternative decision based on the reduced amount of information available from the user, and the system skips The user alternative question 1320 may be asked to construct the asked question 208, which may be a test test and may not affect the final decision 1310.

Referring to FIG. 3, an embodiment of an example image question 1320 that a system can query a user 1314 is shown. In this example, the system can ask the question 1320 of whose answer 1322 can enable the system to better determine the personal characteristics of the user 1314. For example, a question 1320 as illustrated asks, "Which of these causes are interesting?" Where image selection 304 indicates a particular topic, such as pollution, finance, defense, health, and the like. This question 1320 may be targeted to the current user or may be inserted as an empirical question. In an embodiment, the user 1314 is a question 1302 is an empirical question 302 as shown in FIG. 3 with a header that reads "Finally, answer an empirical question submitted by another user." You can be notified.

Referring to FIG. 4, an example embodiment of the type of information 402 that the system may present to the user 1314 when making a particular decision 1310 is presented. For example, the decision 1310 may be for a particular camera, where information about the camera is provided, such as a description, who uses it, the best price for the camera, how to compare with another camera 404, and the like. In embodiments, other determinations 1310 may be provided, such as having a relative ranking 408 such as by score, by percent match, and the like. User 1314 may also be queried for feedback 1312, such as asking if decision 1310 is a good decision. In addition, the user 1314 may have the opportunity to find more information about the decision 1310, such as more information about the product 410, the lowest price finder 412, a website for more advice, and the like. .

Referring to FIGS. 5 and 6, the user 1314 may associate various decisions with respect to the topic as described herein, as presented in connection with the user's request for viewing the parent list, and the like. It may be provided.

In an embodiment, the present invention provides a user with a homepage 700 that includes user 1314 identity 702, personal representation, past decisions, future topics for consideration, 714 decisions made 1310, and so forth. Can be provided to 7 provides an example of a user homepage 700 as the user 1314 sees when they log in to a system account. Here, the system has recommended recent decisions, a list of popular topics 708 to obtain the decision, a search interface 710 to discover the topic, and a status update to the user 1314 who benefits to contribute to the system. There may be indications of recent activity 704, access to the user's profile 712, and the like.

8 and 8A provide an example of a user's profile 712 page displaying information about them and their accounts. The user 1314 may manage user information 802 such as an e-mail address, a password, and the like of the user. They may also answer questions 1320 about themselves, and have memories (810) answers that are automatically used when they use decision making topics in the system. The user 1314 may also receive a reward 804, such as a "badge," and see those displayed as received in response to helping another user, contributing to the system, and the like. Some of these rewards may be based on the quality of the user's contributions, the amount of contributions, and the like. In addition, users may be assigned to demographic groups 808 of people who similarly answer questions 1320 about themselves.

In an embodiment, the user may determine 902 that he or she wishes to contribute expertise to the system, such as in a 'system training' mode. 9 provides, for example, system training for various decisions, rating image quality and user contributed prose, finding duplicate items and questions, contributing new decision making topics, and existing Examples of various links / pages that may allow a user 1314 to contribute, such as contributing a new question 1320 to a topic.

In an embodiment, a user may query a question after selecting a topic for the system to make a decision. 10 provides an example 1000 of how a question 1320 can be presented to a user. As shown, the presentation of the question 1320 to the user 1314 may provide different elements such as topic heading 1002, a picture or illustration 1004 associated with the topic, a set of questions, answer choices, and the like.

After answering the question, the user 1314 may be provided with an answer 1322 or a decision 1310 associated with the user's original question. 11 and 11A illustrate an example of how decision 1310 can be presented to a user 1100 and can include a primary decision, information summarizing the decision, an alternative decision, a modification of the decision, and the like. In addition, the user 1314 may have the opportunity to provide feedback to the system, such as whether the user 1314 agrees with the decision 1310. The user 1314 may also be provided by another proposed topic 1102 based on, for example, the current topic, answer profile, history of answers, user's profile, history of the user's questions, topics that other users find helpful, and the like. Can be.

12 shows an example list of decisions 1200 in a topic. For a product topic as shown, the "determining" may be which product to buy. For other topics, decision 1310 may be "Yes, throw him away" or "No, don't tattoo". The decisions may be ranked and ordered based on their relevance to the user, based on how the user 1314 answered the question, based on how the user 1314 answered the question 1320 of the topic, and the like. In addition, items may be ranked by price, name, or the like.

16 shows an example of a contributor / expert interface showing recent contributions to system 1602 and other users who have made contributions 1604. The upper right corner is a question for learning a user's preferences 1608.

FIG. 17 shows an example of an objective question 1700, in this case an interactive question with a system that inquires what the user is looking for to help determine which dog is their dog.

18 is ranked as by the specific recommended decision 1800 (in this case, you name the Rusty), a review 1802 of this decision from another user (by their similarity to the user). Can be used), a Yes / No button 1804 (such as receiving feedback on this decision to indicate another decision area for the user to enjoy), and an example of a decision result representing the proposed topic 1808. In this example, the second and third best recommended decisions of the system are listed under # 2 tab 1810 and # 3 tab 1812. The system may also couple to the search by recommending a "wild card" decision, which may be a decision taken partially through randomization. The proposed topic 1808 is selected based on how appropriately the system considers these topics may be for the user and / or how much the system may be able to generate from the user using these different decision areas. Can be.

19 shows an example of an interface for a user 1900 to establish an association between an attribute and a decision result. In this example, the determination result "Rusty" should be associated with the attribute "Is this name for female or male dogs? Male".

20 illustrates an example of how a user can edit content in system 2000. In this example, the user can edit the decision result, its name, description, URL to get more information, and the like.

FIG. 21 shows an example of how content that is editable by a user may also have an interface for viewing a previous modification 2100 for the content and indicating a change between two modifications. The user can also revert changes made by other users if these changes are considered irrelevant or unhelpful. In this case, the example shows the difference between the two modifications to the decision result whose description of the result has changed.

22 shows an example showing a question edited by the user 2200. New answers can be added, existing answers reordered, question and answer text itself edited, and so on. The question may optionally be "locked" to prevent other users from changing them, as indicated by the pad lock icon 2202.

FIG. 23 shows an example indicating that an edit to an attribute may have the same modification history as other editable content 2300. This example shows the difference between two modifications of the attribute association between the decision result "Rusty" and the attribute "How many syllables do you want to have? Less than two or three or more is good".

FIG. 24 shows an example of displaying a 'workshop' screen 2400 where a newly added area of advice can be displayed first. In an embodiment, the expert user may add to the regular user without seeing the work in progress. Content may be objectionable, and content that is considered irrelevant or of low quality may be voted off and removed from the system.

25 shows an example showing that the system queries the user preference / subject question 2500 to learn preference and subjective preference from the user. After answering these questions, the system can display statistics about how other users answered the same question.

FIG. 26 shows an example of an activity feed 2600 of recent activity by contributors across sites displaying newly added content and expert training.

In an embodiment, the present invention is an improved manner to provide a decision to the user 1314 with questions 1320 across a broad category of topics including product, individual, health, business, politics, education, entertainment, environment, and the like. It can provide the equipment to provide. For example, the system may be responsible for everything from whether a user 1314 breaks up with their boyfriend, whether you need a tattoo, whether you can deduct something from your tax in addition to product decisions, and so on. May provide a decision. In an embodiment, the system can provide a determination of any interest the user 1314 may have.

In an embodiment, the present invention may provide a decision system that is resilient and capable of change and growth. This may be partially enabled by the use of the system of the question 1320 and the dialogue of answers to make a decision, followed by obtaining feedback from the user 1314, thus improving the system. In an embodiment, this approach can be considerably more powerful because the system can ask any question 1320 and thus get much better information from the user 1314 about what they want. In addition, the user may be able to expand the system by entering their own questions 1320 and answers for the system for inquiries, entering new decisions for the system to make, and the like. The system may then automatically try the newly entered information for the baby whether it is useful or helpful, use this new information to determine if it is useful, and possibly not helpful to the user. You can stop asking / using questions / decisions. In an embodiment, this approach may potentially provide for building a wisdom-based decision of a crowd that manufactures an expert system for any topic.

In an embodiment, the present invention may also provide improved decision equipment to the user 1314 by providing decisions by ranking across non-traditional feature dimensions. For example, instead of just ranking cameras by price or size, the system can rank cameras based on how much they like by retired people, how sexy they look, and so on. The system can then help the user navigate across these dimensions. For example, instead of enabling the user to say "I like this camera but want a cheaper one", the system says "I like this camera but I want something better to learn photography" or "I I like vacations, but I want to have a more active social scene. ”

In embodiments, the present invention is directed to a variety of different technologies, such as web interfaces, instant messaging, voice, cell phones, SMS / instant messaging, third party use (eg, widgets on third parties, web services sold to third parties), and the like. It can contribute to the user interface. For example, a voice interface may be well suited to a system because there may be a very limited vocabulary that the system must recognize, such as a possible answer to each question. In this way, if the system cannot understand the user responses, it may move to another question 1320 instead of annoying the user 1314 by contacting them to continue repeating their answers. In another example, the present invention may be incorporated into a third party site, such as a search of a TV on an e-commerce website, where the present invention is a widget to help the user 1314 narrow the results or is a good choice for them. Use the present invention as a widget in connection with a real estate website to build an MLS query to the user 1314 to find a matching home. In an embodiment, the present invention may provide a user interface, wherein the decision to provide the user 1314 with a significantly improved approach to the physical interface and to obtain decisions on various topics both in the manner of questions and answers. do.

In an embodiment, the present invention may be incorporated into third party products in this manner to enhance third party user interfaces and user satisfaction. For example, some website services provide predictions through past purchase history. In this case, the present invention may be able to explore the mood or intention of the user, such as by asking an explicit question. In the case of a search engine, the present invention may detect when a user 1314 makes a decision 1310 and then begin to query them according to a question. In the case of forum sites, mailing lists, newsgroups, etc., the present invention can provide improved access to decisions and decisions made by people similar to the user. For example, the present invention may be able to find people who are in the same situation as the user and search through all forum posts by providing which decision 1310 is to recommend the forum community to them.

In an embodiment, the present invention may be capable of extending an e-commerce web application user interface. For example, user 1314 may begin a product search with a keyword search and then ask question 1320 to narrow the results to the best decision 1310 for the user. The present invention may be able to provide a Q & A interface for retrieving products once the user 1314 clicks on the category page. For example, after clicking on a camera on a website, user 1314 may see the first question. The present invention may be able to rank products along a dimension based on how the user answers the question. For example, the camera answers the question “What do you want the camera for?” How do people answer 1322 “Travel” and the worst from the best based on whether they provide positive or negative feedback for a particular camera It can be ranked as a 'travel camera'. This allows the e-commerce website to rank a list of camera keyword search results from best to worst travel cameras.

In an embodiment, the present invention may be capable of providing enhanced search engine capabilities, such as detecting when a user 1314 attempts to make a decision 1310 and switching to a Q & A interface, resulting in a search result from a keyword search. Ask follow-up questions 1320 to narrow or rerank the results based on, query questions 1320 to build keyword search queries, or refine search queries, and which links to users after the questions are asked. Learning feedback based on whether 1314 clicks or the like. In addition, the present invention may implicitly learn about the user 1314 and include those such as when they use the system, where they are in the world, what type of browser they are using, the weather where they are, and the like. You can change the ranking based on implicit facts.

In an embodiment, the present invention may be capable of providing a manner in which information is collected and used by a user. For example, Wikipedia is a way for a user to contribute information so that the end user 1314 must self-validate the accuracy of the information subsequently supplied to them to some extent. In a similar manner, the present invention may be capable of hosting web applications that utilize user-contributed content. For example, instead of learning what the price of the camera is, the web application can allow the user to enter the price of the camera and then allow other users to self-validate these claims. In this way, the categories of contributed information can be allowed to set organically as the user interacts with the system.

In an embodiment, some e-commerce applications may provide products and / or services related to personal preferences, and thus may benefit from the present invention. For example, there are currently a number of movie rental web services, where user 1314 selects a movie for delivery to their home via mail. The decision is also provided to the user 1314 based on what the user 1314 has selected in the past. However, choosing a movie can involve personal interest in rental that cannot be determined by mood, intent, weather, whether they are alone or with whom, their current personal relationships, and the like. These types of interest can be explored by the present invention through questions, thus providing a much more personalized match to the user's interest in rental.

In an embodiment, the local search application may be enhanced through the use of the present invention. For example, if the user 1314 wants a decision 1310 about where to have dinner, they can search for "evening in New York" and find a website with a suggestion targeted to the query. However, this interface is insufficient when the user 1314 does not have a clear idea of which keywords should be included. For example, user 1314 may not be aware of key options for food, but may not be thinking about searching for 'Ethiopian food New York'. The present invention may have the advantage that it may be possible to determine which question 1320 should be queried to narrow the possibilities. In an embodiment, the present invention may be able to assist in the construction of a search query.

In embodiments, the present invention may provide an improved way of matching users and professionals, users, and other knowledge-based users. For example, services may be provided to collect users and experts at different topics. The user can then enter the Q & A session, which comes to the service's web interface and the best match is determined. As a result of the question, the system may provide a decision, where a profile of an expert or other user 1314 is provided and the user 1314 may be queried if they agree with the individual recommended. In an embodiment, the user 1314 may be provided on a home page where previous matches and communications are maintained, forwarded to friends, rated by experts, and the like.

In an embodiment, the present invention may provide a platform for community based question 1320 and answer 1322 applications. For example, a user may post a question 1320 to the system and another user may be allowed to respond. In such a system, user 1314 may receive answers from a single user, multi-user, automated system, where user 1314 may be able to select which answers 1322 feel they are correct. This answer 1322 may be private, posted for others to view, posted as an accurate answer, provided to the system, and the like. In an embodiment, the system may use questions 1320 and answers to further develop the system, provide more accurate answers to the user, classify the answers provided to the user, filter the answers provided to the user, and the like. In addition, a user of the system can contribute feedback to the answers provided by other users and contribute filtering criteria to eliminate inaccurate answers.

In an embodiment, the present invention may be used as entertainment through machine learning capabilities as described herein. For example, user 1314 may provide input or thoughts of thought, such as topics, keywords, categories, questions, feelings, etc., and the system may make inferences about through a series of questions 1320 and answers. . For example, user 1314 may consider an object, such as baseball, and the system may utilize machine learning capabilities, such as a geometric system, to provide the user with questions 1320. Common question 1320 may relate to a size such as 'Is this larger than a toast?' This question 1320 can then be answered by the user, such as through multiple selection, filling in the blanks, through true / false, through free responses, and the like. The system may then continue the question 1320 and answer 1322 sequence until it has a guess, and provide this guess to the user. In an embodiment, this process may continue for a fixed number of questions, a number of random questions, a number of user 1314 assigned questions, a number of system determined questions, a number of system specified questions, and the like. In an embodiment, the system provides a user interface via the Internet via a website, via a self-contained computing device, via a mobile computing device, via a telephone service, via a voice interface, in connection with an instant messaging service, via text messaging. May be provided to the user 1314. In an embodiment, the system may be provided to a third party, such as as a widget to another website, as an API to a third party application, and so on. In embodiments, the present invention may use a non-neutral network for entertainment applications such as playing games.

In an embodiment, the invention may provide a system to assist in the discovery of a new drug, wherein the system may provide assistance in the selection and combination of molecules in the production of the new drug. For example, the system may ask the user 1314 for information related to chemical parameters such as solubility, reactivity, toxicity, etc., and ask questions 1320 and these to explore the user's expertise in recognizing molecular structure. Can be combined. As the question 1320 and answer 1322 sequence proceeds, the system can provide the user 1314 with insight into which molecular structure can be stable and synthesizable. In an embodiment, the process may result in a new potential drug until the user 1314 has an improved concept of which molecular combinations can be made for the new drug, until the selection of a new inquiry route is available for presentation to the user. Until it is identified and so on.

In an embodiment, the present invention may provide an image finder application, where the user 1314 may assist in identifying images that conform to some subjective criteria that are not necessarily explicitly recognized by the user. For example, user 1314 may be involved in the development of a brochure for a company, where they have text for the brochure, but must select an image to support the thoughts and feelings the text attempts to convey. The user 1314 may have a subjective idea of what type of picture may be required in this case, but not necessarily to the extent that they can specify a search by keyword. The user 1314 may instead first specify the source of the image, such as from a file, database, website service, from a Google image, or from an advertiser image bank. The user 1314 may then be asked a series of questions or presented therefrom with a series of images for selection. The answer and / or selection that the user 1314 chooses may then be used to improve the selection presented to the user, from which further questions 1320 and / or image selection may be provided. In an embodiment, this process may continue until the user 1314 finds the image for selection as the final image. Additionally, the system can take the user's 'final selection' and select another group of similar images for presentation to the user at a time at which the user 1314 can choose to continue the process of improving the selection.

In embodiments, the present invention may be used in baby naming applications, where user 1314 may only have a vague concept of what their preferred names are. The user 1314 may provide information to the system to assist in the learning of the user's preferences, such as questions 1320 about family, friends, education, heritage, geographic location, place of birth, hobbies, books read, movies watched, etc. Different types of questions 1320, as intended, may be initially asked. The system may then continue to learn through the presentation of questions 1320 related to name preferences in a number of ways, such as name ratings, selection from a list of names, answering questions 1320 belonging to the name, and the like. In an embodiment, this process may continue until the user 1314 finds a name to select as the last name. Additionally, the system can take the user's 'final selection' and select another similarly named group for presentation to the user at a time at which the user 1314 can choose to continue the process of improving the selection.

In embodiments, the invention is not limited to these, but video games, laptops, vacations, cameras, general personalities, beverages, cell phones, televisions, grills, watches, coffee machines, toelets, Halloween, GPS devices, sexiest entertainers , Your personal heroes, presidential elections, baby toys, blogs, camcorders, cars, which Star Wars characters are you, credit cards, hair care, skin care, sex and the city, tattoos, job, how much, City of residence, Dog breed, Fragrance, New York, Neighborhood selector, Software, Desktop computer, DVD player and recorder, Cigarette, Mercy, Broadware show, Speaker, Home theater system, MP3 player, Computer networking device, Headphones, Memory Cards, Magazines, Books, Oprah Fixes, Books, New York Times Best Sellers, Business Casual Clothing, Franchise, Cooking Utensils, Toys, Difficulty-education, sportswear, espresso machine, should go to Greece, should be independent from parents, should raise wages, have drinking problems, should ADD / ADHD be given medication, vacuum cleaner, washer and dryer, newborn Is the company right for me, the hygrometer, are you a good friend, the risk of developing diabetes, which language to learn, microwave, car audio, what kind of customer, wine, the military should apply, which military division Should I, what kind of art to enjoy, kids and car seats, baby strollers, baby travel accessories, natural and organic beauty products, makeup, home audio receivers and amplifiers, copiers and fax machines, printers, boyfriends / girlfriends Breaking up, which Greek gods are you, which games you will play, computer accessories, which superpowers you should have, universities, Line degree program, choosing a major in college, preventing identity theft, hiring a personal trainer, buying a car or leasing, doing laser eye surgery, what to do about hair loss, starting my business, Should I send my children to kindergarten, how should I treat my family when I visit NYC, need OTC pain relief, willingness to live, mileage or cash for the next flight, best way to whiten teeth, make up my daughter Should I do it, is it possible hypnosis to fix my bad habits, ED options, sleep aids, OTC allergy pills, how much should I spend on wedding gifts, should I buy an extended warranty, or would it be better to get an SAT or ACT? Personal Audio Accessories, Enjoy Coffee / Espresso Drinks, Video Game Consoles, Jeans, Downloadable PC Games, Snacks, Vita And supplements, which superhero I am, sunglasses, kitchen utensils, pillows, beauty accessories, beauty bags and cases, sporting goods, which instruments are right for me, must hire a decorator, a leading leader, belonging to a mall, power washer , Small business, phone system, how much tip, should Botox, liposuction, skin cancer risk, should refinance my home, vehicle service (NYC), micro brewed beer, gourmet chocolate, I am retired Save Enough, Entertainment Centers / TV Stands, Cookbooks, Electric Shavers, Send Nephew / Near Baby Gifts, Luggage, Computer Projectors, Energy / Gymnastics Bars, Razors, Gourmet Ice Cream, Online Dating, Newscasts, Makeup, Tools And brushes, beauty mirrors and compacts, business books, how quickly to call after your first date, place to retire, etc. Should you sell hard drives, universal remotes, walking shoes, life insurance policies, how much do you think of the environment, eating disorders, baby beds, diet and diet books, cell phone plans, wedding and engagement rings, stubborn enough, my kids Play too much video games, tax preparation (personal refunds), reverse mortgages, cancel plans with friends on dates, children's TV shows, kitchen counters, bathroom supplies, pest control, cancer specialists, hospitals, nationwide Chain restaurants, cereals, should I have a child now, should I hire a nanny, movies, beef sites, daily goal calories, OCD, home air purifiers, automatic air purifiers and purifiers, iPhone applications, gay / lesbian vacations, my colleagues Can I ask for a date? Is my child just about puberty ready to take care of my baby? Energy drinks, TV shows, office furniture, motorcycles, children's rewards for good report cards, lawn trimmers and hacksaws, stressful, religion, make a good first impression, spend too much time online, new hair Should I style, should my children be home schooled, should I use diaper bags, cloth diapers or disposable diapers, dog toys, my partner cheats on me, classic books, my old parents should stop driving, Is it over with my former spouse, is this desire or love, pedometer and heart rate monitor, chewing gum, weather device, gas additives will help my car, Orlando theme park, how big Turkey to buy, popular music-new releases, Magnetic Tanning Machine, Tax & Money Management, Software, Baby Bottle & Sippy Cup, Baby Height Chair & Booster Seat Savings for retirement, baby bands, toast and toaster ovens, comfort sheets and bedspreads, dish sets, pet carriers and kennels, cheeses, kitchen faucets, casual shoes, dress shoes, beauty electronics, retirement, mutual fund selectors, Steak area, what is my D & D arrangement, acne and acne medication, bathroom faucet, exterior lighting, landscape lighting, lawn mowers, aperitif, cognac, rum, scotch, tequila vodka, whiskey, las vegas show, sunscreen, running shoes, US MBA program, patio and outdoor furniture, kitchen knives, true fan, car insurance, personal legal services, financial advisors should be hired, indoor plant selector, delivery service, this must be deducted, full heater, sofa, house number, Contact lenses, birthday gifts, my career is the best, ebooks, door handles and lock sets, snow removal equipment, green home improvements, children's clothing and clothing Clothing, motorcycle helmets, bicycle helmets, juicers, golf clubs, refrigerators, wine coolers, ranges and ovens, air conditioners, Christmas gifts, farewell gear, rash medications, diabetes monitoring devices, smoking aids, how to do my hair on my back, menopause neutralization Hormones, Hiking Backpacks, School Backpacks, Website / Domain Acquisition, Email Service, Web Hosting, Carpets, Power Tools, Tiles, Water Heaters, Outdoor Paint, Window Treatment, Stove Screen, Indoor Lamps, Small Business Legal Services, Brunch Recipes, Ceilings Fans, Mattresses, Las Vegas Hotels and Casinos, Salsa, Love Quiz for Valentine's Day, How much to spend on client gifts, Anniversary Gifts, Outdoor Coats, Casual Coats, Camping Tents, Sleeping Bags, Tires, Adventure Vacations, Music Downloads, Video Downloads, Weddings Dresses, Wedding Themes, Manhattan Gyms, Budget Hotel Chains, Golf Courses, Ski Vacations, USA Wave, ETF fund, designer handbag, bankruptcy one, 401k down payment at home, psychiatrist one, self defense, plates, dishwasher, political group, new year resolutions, cruise line, family vacation, baby food, baby health care Products, need to shave hair, t-shirts, online photo services, buy class graduation rings, summer job / internships, where to volunteer, home alarm system, diagnose your relationship issues, should she / he appeal to me or adopt , My old parents should drive, online bank account, BBQ sauce, frozen pizza, recipe finder, one to present again, bodybuilding supplements, household exercise equipment, several hours sleep, one should consider plastic surgery, arthritis risk, Risk of heart disease, risk of osteoporosis, gambling problems, best dances, bikes, cat food, dog food, hobby picks, martial arts, Other must-haves, which muffs, activities for children, how moral, should my children be baptized, Miami hotels, US national parks, motor oil, car videos, blouses, coats, dresses, frames, socks, interview clothing, jackets , Dressing gown, pants, shirt, skirt, hat, phone-home phone, steakhouse, which birth method is right for me, summer camp recommender, frenzy bracket picker in March, infant formula, New York bakery, partial jet ownership, how confident Is there a digital photo frame, accountant needed, my children have ADD / ADHD, a decision on a plurality of topics including document shredders, baby monitors, green home improvements, conference calls, and the like.

In an embodiment, as shown in FIG. 13, the present invention may help the user 1314 make the decision 1310 through the use of the machine learning equipment 1302. The process may begin with an initial question 1320, received 1304 by the machine learning equipment 1318 from the user 1314. User 1314 may then have a conversation 1308 consisting of questions 1320 from machine learning equipment 1318 and answers 1322 provided by user 1314. Machine learning equipment 1318 may then belong to initial questions 1304, such as recommendations, diagnoses, conclusions, advice, and the like, and provide a decision 1310 to the user based on conversation 1308. In an embodiment, future questions 1320 and decisions 1310 provided by machine learning equipment 1318 may be enhanced through feedback 1312 provided by user 1314.

In an embodiment, the initial question 1304 given by the user 1314 may be an objective question, a subjective question, or the like. The question 1320 may be provided from a wide range of topics such as topics belonging to product, personal information, personal health, economic health, business, politics, education, entertainment, environment, and the like. The question 1320 may be in the form of a multiple choice question, a yes-no question, a rating, a selection of an image, a private question, or the like. The question 1320 may be for the user 1314, may be provided by another user, may be provided by an expert, or the like. The question 1320 may be previously answered, for example, from a current conversation 1308 with the user 1314, from a saved previous conversation 1308 with the user 1314, or from a previously saved conversation 1308 with another user. Can be based on answers. The question 1320 may be a pseudo-random question, such as a test question, a search question 1320 that helps select the pseudo-random decision 1310 at an opportunity that the pseudo-random decision 1310 turns out to be useful. The question 1320 may include at least one image as part of the question. Question 1320 may follow a psychographic dimension. In an embodiment, the question 1320 may not contact the user 1314 directly, but rather relate to an IP address, the user's location, the weather at the user's location, domain name, route information, relevance to recent downloads, recent From conceptual information, such as through relevance to network access, relevance to recent file access, and so forth.

In an embodiment, the conversation 1308 continues until the machine learning equipment 1318 develops high confidence in the set of reduced decisions presented to the user, such as a single decision 1310 presented to the user. Can be. The decision 1310 provided by the machine learning equipment 1318 may be independent of the order of the questions in the conversation 1308. The decision 1310 can provide an alternative decision 1310 when at least one question 1320 of the conversation is omitted, where the alternative decision 1310 has less information from the user 1314. ) May be different. Decision 1310 may indicate a ranking of decision choices, such as ranking decisions across non-traditional feature dimensions. Decision 1310 can display at least one image related to decision 1310. Decision 1310 may be a pseudorandom decision about an opportunity for which pseudorandom decision 1310 turns out to be useful, such as the pseudorandom decision being part of a system of search, where the system of search may improve the effectiveness of the system and Machine learning equipment 1318 may be learned from the search.

In an embodiment, the provided feedback 1312 how the user 1314 answers the question 1320 in the conversation 1308, and how the user 1314 responds to the decision 1310 provided by the machine learning equipment 1318. Or derived from or the like. In an embodiment, feedback 1312 may be solicited from user 1314.

In an embodiment, user 1314 may extend the learning of machine learning equipment 1318 by entering new information, where the new information may be their own topics, questions, answers, decisions, and the like. Machine learning equipment 1318 may use the new information to determine whether the new information is helpful to the user.

In embodiments, the user interface may be provided for user interaction with machine learning equipment 1318 such as those associated with web interfaces, instant messaging, voice interfaces, mobile phones, SMS, and the like.

In an embodiment, the present invention can help a user make decision 1310 through the use of machine learning equipment 1318. The process may begin with an initial question 1304 received from the user 1314 by the machine learning equipment 1318, where the initial question 1304 may be a product, individual, health, business, politics, education, entertainment, environment, or the like. It may be associated with one of a broad category of the same topic. User 1314 may then have a conversation 1308 consisting of questions 1320 from machine learning equipment 1318 and answers 1322 provided by user 1314. Machine learning equipment 1318 may then belong to initial questions 1304, such as recommendations, diagnoses, conclusions, advice, and the like, and provide a decision 1310 to the user 1314 based on the conversation 1308. In an embodiment, future questions 1320 and decisions 1310 provided by machine learning equipment 1318 may be enhanced through feedback 1312 provided by user 1314.

In an embodiment, as shown in FIG. 14, the present invention may assist a user in making a decision 1310 through the use of computing equipment 1402. The process may begin with an initial query 1304 received by the computing equipment 1418 from the user 1314. The user 1314 may then have a conversation 1408 consisting of a question 1320 from the computing equipment 1418 and an answer 1322 provided by the user 1314. The computing equipment 1418 may then provide a decision 1310 to the user 1314 based on the feedback 1428 collected from the feedback from the plurality of users 1412. In an embodiment, computing equipment 1418 may enhance future questions 1320 and decisions 1310 provided by computing equipment 1418 based on received feedback 1412 from the user.

In an embodiment, the present invention can help a user make decision 1310 through the use of machine learning equipment 1318. The process may begin with an initial question 1304 received by the machine learning equipment 1318 from the user 1314. The user 1314 may then have a conversation 1308 consisting of questions 1320 from the machine learning equipment 1318 and answers 1322 provided by the user 1314, where the conversation 1308 is provided. The number of questions 1320 and answers 1322 can determine the quality of decision 1310. Machine learning equipment 1318 may then belong to initial questions 1304, such as recommendations, diagnoses, conclusions, advice, and the like, and provide a decision 1310 to the user 1314 based on the conversation 1308. In an embodiment, future questions 1320 and decisions 1310 provided by machine learning equipment 1318 may be enhanced through feedback 1312 provided by the user. In an embodiment, the quality may be high when the number of questions 1320 and answers 1322 is large, such as more than 10 questions, more than 15 questions, more than 10 questions, and the like. In an embodiment, the quality may be good quality when the number of questions 1320 and answers 1322 is small, such as less than 10 questions, less than 5 questions, less than 3 questions, one question, and the like.

In an embodiment, as shown in FIG. 15, the present invention may make decision 1310 through the use of machine learning equipment 1318. The system is machine learning equipment 1318 capable of receiving an initial question 1304 from a user 1314, machine learning equipment 1318 providing a question 1320 to a user 1314 and receiving an answer 1322 from the user. Conversation equipment 1502, machine learning equipment 1318 that provides decision 1310 from decision equipment 1504 to user 1314, and the like. In an embodiment, the decision 1310 provided to the user 1314 is based on the exchange of conversation 1308 between the user 1314 and the machine learning equipment 1318 and may belong to the initial question 1304. In addition, the machine learning equipment 1318 receives feedback 1312 from the user 1314 through the feedback equipment 1508 from the user 1314 to improve future questions 1320 and decisions 1310 provided by the machine learning equipment 1318. Can be received.

In an embodiment, the present invention may help the user 1314 make the decision 1310 through the use of the machine learning equipment 1318. The process may begin with an initial question 1304 received by the machine learning equipment 1318 from the user 1314 via a third party, such as a search application, social network application, service provider, comparative shopping engine, media company's web environment, and the like. have. User 1314 may then have a conversation 1308 consisting of questions 1320 from machine learning equipment 1318 and answers 1322 provided by user 1314. Machine learning equipment 1318 may then belong to initial questions 1304, such as recommendations, diagnoses, conclusions, advice, and the like, and provide a decision 1310 to the user 1314 based on the conversation 1308. In an embodiment, future questions 1320 and decisions 1310 provided by machine learning equipment 1318 may be enhanced through feedback 1312 provided by user 1314.

In an embodiment, the present invention may help the user 1314 make the decision 1310 through the use of the machine learning equipment 1318. The process may begin with an initial question 1304 received by the machine learning equipment 1318 from the user 1314 via a third party search application, where the user starts with a keyword search on the third party search application and then the machine A conversation 1308 is provided that consists of a question 1320 from learning equipment 1318 and an answer 1322 provided by user 1314. Machine learning equipment 1318 may then belong to initial question 1304 and provide decision 1310 to user 1314 based on conversation 1308, where decision 1310 may be in the form of a sorted list. As well as a third party search application.

In an embodiment, the present invention may help the user 1314 make the decision 1310 through the use of the machine learning equipment 1318. The process may begin with an initial question 1304 received by the machine learning equipment 1318 from the user 1314. The user 1314 may then have a conversation 1308 consisting of questions 1320 from the machine learning equipment 1318 and answers 1322 provided by the user 1314, where the machine learning equipment 1318 is Third party information, functions, utilities, etc. may be available. Machine learning equipment 1318 may then belong to initial questions 1304, such as recommendations, diagnoses, conclusions, advice, and the like, and provide a decision 1310 to the user 1314 based on the conversation 1308. In embodiments, third party information, functions, utilities, and the like may include an application programming interface (API) that enables the collection of cost information, product information, personal information, topic information, and the like.

In an embodiment, the present invention may help the user 1314 make the decision 1310 through the use of the machine learning equipment 1318. The process may begin with an initial question 1304 received by the machine learning equipment 1318 from the user 1314 via a third party search application, where the user 1314 is a keyword search on the third party search application. A conversation 1308 is provided that begins and then consists of a question 1320 from the machine learning equipment 1318 and an answer 1322 provided by the user 1314. Machine learning equipment 1318 may then belong to initial questions 1304, such as recommendations, diagnoses, conclusions, advice, and the like, and provide a decision 1310 to the user 1314 based on the conversation 1308. In an embodiment, the decision 1310 may be provided again to the third party search application based at least in part on the collaborative filtering.

In an embodiment, the present invention may help the user 1314 make the decision 1310 through the use of the machine learning equipment 1318. The process may begin with an initial question 1304 received by the machine learning equipment 1318 from the user 1314. User 1314 may then have a conversation 1308 consisting of questions 1320 from machine learning equipment 1318 and answers 1322 provided by user 1314. Machine learning equipment 1318 may then belong to initial questions 1304, such as recommendations, diagnoses, conclusions, advice, and the like, and provide at least one image decision 1310 to the user 1314 based on the conversation 1308. . In an embodiment, the image may be a photo, picture, video image, advertisement, or the like.

In an embodiment, the present invention may help the user 1314 make the decision 1310 through the use of the machine learning equipment 1318. The process may begin with an initial question 1304 received by the machine learning equipment 1318 from the user 1314. User 1314 may then have a conversation 1308 consisting of questions 1320 from machine learning equipment 1318 and answers 1322 provided by user 1314. Machine learning equipment 1318 may then belong to initial questions 1304, such as recommendations, diagnoses, conclusions, advice, and the like, and provide a decision 1310 to the user 1314 based on the conversation 1308. In an embodiment, the decision 1310 may be based at least in part on learning from the decision 1310 provided by another user of the machine learning equipment 1318.

In an embodiment, the present invention may help the user 1314 make the decision 1310 through the use of the machine learning equipment 1318. The process may begin with an initial question 1304 received by the machine learning equipment 1318 from the user 1314. User 1314 may then have a conversation 1308 consisting of questions 1320 from machine learning equipment 1318 and answers 1322 provided by user 1314. Machine learning equipment 1318 may then belong to initial questions 1304, such as recommendations, diagnoses, conclusions, advice, and the like, and provide a decision 1310 to the user 1314 based on the conversation 1308. In an embodiment, the decision 1310 may be based at least in part on collaborative filtering.

In an embodiment, the present invention may help the user 1314 make the decision 1310 through the use of the machine learning equipment 1318. The process may begin with an initial question 1304 received by the machine learning equipment 1318 from the user 1314. User 1314 may then have a conversation 1308 consisting of questions 1320 from machine learning equipment 1318 and answers 1322 provided by user 1314. Machine learning equipment 1318 may then belong to initial questions 1304, such as recommendations, diagnoses, conclusions, advice, and the like, and provide a decision 1310 to the user 1314 based on the conversation 1308. In an embodiment, the decision 1310 may be based at least in part on collaborative filtering in which the context is provided via conversation 1308, such as at least one question providing a context for collaborative filtering.

In an embodiment, the present invention may help the user 1314 make the decision 1310 through the use of the machine learning equipment 1318. The process may begin with an initial question 1304 received by the machine learning equipment 1318 from the user 1314. User 1314 may then have a conversation 1308 consisting of questions 1320 from machine learning equipment 1318 and answers 1322 provided by user 1314. Machine learning equipment 1318 may then belong to initial questions 1304, such as recommendations, diagnoses, conclusions, advice, and the like, and provide a decision 1310 to the user 1314 based on the conversation 1308. In an embodiment, the decision 1310 may be based solely on information collected belonging to the initial question 1304 and collected through the plurality of users 1314 of the machine learning equipment 1318, wherein the decision of the machine learning equipment 1318 is performed. At least one of the plurality of users 1314 may be a user 1314 associated with the conversation 1308.

In an embodiment, the present invention may help the user 1314 make the decision 1310 through the use of the machine learning equipment 1318. The process may begin with an initial question 1304 received by the machine learning equipment 1318 from the user 1314. User 1314 may then have a conversation 1308 consisting of questions 1320 from machine learning equipment 1318 and answers 1322 provided by user 1314. The machine learning equipment 1318 then belongs to the initial question 1304 and is determined to the user 1314 based on the dialogue 1308 and with limited initial machine learning equipment 1318 knowledge of the gist of the initial question 1304. 1310 may be provided. In an embodiment, the limited initial machine learning equipment 1318 knowledge may be seed knowledge, may be limited to basic knowledge related to the subject matter of the initial question 1304, and the initial question 1304, where the basic knowledge may be expert knowledge. May be limited to basic knowledge related to the subject matter.

In an embodiment, the present invention may help the user 1314 make the decision 1310 through the use of the machine learning equipment 1318. The process may begin with an initial question 1304 received by the machine learning equipment 1318 from the user 1314. User 1314 may then have a conversation 1308 consisting of questions 1320 from machine learning equipment 1318 and answers 1322 provided by user 1314. Machine learning equipment 1318 may then belong to initial questions 1304, such as recommendations, diagnoses, conclusions, advice, and the like, and provide decision 1310 to user 1314 based on conversation 1308, and decision 1310. ) May be based on learning from a combination of expert and user input.

In an embodiment, the present invention may help the user 1314 make the decision 1310 through the use of the machine learning equipment 1318. The process may begin with an initial question 1304 received by the machine learning equipment 1318 from the user 1314. User 1314 may then have a conversation 1308 consisting of questions 1320 from machine learning equipment 1318 and answers 1322 provided by user 1314. Machine learning equipment 1318 may then belong to initial questions 1304, such as recommendations, diagnoses, conclusions, advice, and the like, and provide category-based decisions 1310 to user 1314 based on conversation 1308.

In an embodiment, the present invention may help the user 1314 make the decision 1310 through the use of the machine learning equipment 1318. The process may begin with an initial question 1304 received by the machine learning equipment 1318 from the user 1314. User 1314 may then have a conversation 1308 consisting of questions 1320 from machine learning equipment 1318 and answers 1322 provided by user 1314. Machine learning equipment 1318 may then be retrieved from a plurality of users 1314 of machine learning equipment 1318 for machine learning equipment 1318 to classify and provide a decision 1310 along at least one of the psychographic and demographic dimensions. A decision 1310 can be provided to the user 1314 who can use the response.

In an embodiment, the present invention may provide a response to the user 1314 through the use of the machine learning equipment 1318. The user 1314 may have a conversation 1308 consisting of a question 1320 from the machine learning equipment 1318 and an answer 1322 provided by the user 1314, where from the machine learning equipment 1318 The question 1320 may relate to applications such as entertainment applications, drug delivery applications, baby name applications, and the like. Machine learning equipment 1318 may then provide a response to user 1314 based on conversation 1308 belonging to initial question 1304 such as recommendation, diagnosis, conclusion, advice, and the like. In an embodiment, future questions 1320 and decisions 1310 provided by machine learning equipment 1318 may be enhanced through feedback 1312 provided by user 1314.

In embodiments, the present invention may provide results based on multiple dimensions, such as results based on text matching from user input based on a user's preference profile, and the like. 27 shows an example search interface showing a search result for a user query "pios". In this example, the ranking is based on a first discovery decision and decision result (recommended), which is a good text match for the user query, and then a second ranking of decision result by knowledge that the system has about the user's preference profile. In this example, "Pios" is a good text match for "Verizon Pios" determined as a recommendation to "Which ISP should I use" (2702) and "Which US satellite / cable service should I get?" (2704). , Both are ranked # 1 of this user based on the user's preferences.

In an embodiment, the present invention may provide a plurality of question and answer 'results' in response to the user's unstructured input. For example, FIG. 28 illustrates a case where a user query is 'suv'. The first result is for the topic "Which new car should I buy?" (2802) with the question 'Which type of car do you want to buy?' (2802) already answered with 'suv'. In addition, as shown in the example, the top three results of the user are personalized and displayed based on their preference profile. For example, "Which BMW should I buy?" (2804), "Which new model should I buy?" (2808), and "Which tires should I use for my car or truck?" (2810) Through traditional Q & A interfaces, such as for a question, you do a keyword search and get results without explicitly answering any question.

In embodiments, the present invention may inquire of the user to express their preferences when they provide answers to questions that contradict each other and are mutually exclusive or each have a dramatic impact on the ranking of results for the user individually. For example, a user may start the topic "Which new car should I buy" and answer that they want to buy an SUV for less than $ 18,000 and are more practical than luxury. The system may want to get the user's preference for something more important-the car is an SUV or less than $ 18,000. 29 shows an example question to the user asking for their preferences.

In an embodiment, the system may learn a set of question importance, per user, per decision result, per question-answer-decision result, and the like. For example, the system can learn that User A cares more about weight than price when he chooses to buy a small camera, but when he buys a car he cares more about price than others. have.

In an embodiment, the user can expand the system by adding new decision results. These new results may optionally include a link to a web page to read more about the decision result. The system can automatically switch these links so that the system associates the links so that the system receives a referral from the site to which the link points. Also, based on the link submitted by the user, the system can recognize what kind of link it is, and how to analyze information such as the price for the product, such as from Amazon.com, or analyze the product code to determine merchant specific APIs. It can be appreciated that the call can be made to find product information based on the product code.

In embodiments, users may be clustered into groups using dimensionality reduction techniques such as singular value decomposition (SVD), eigenvectors, and other similar based approaches. The system can display information about why groups of users are clustered together. One way to do this is to find the upper X dimension in the lower dimension space where the cluster is most different from the population mean as a whole. The divergence of the distribution of clusters of answers from the distribution of the general population in each dimension in the subspace can be used to rank the dimensions in terms of how well they describe what is unique for each cluster.

In an embodiment, the dimensions of the subspace cannot be easily explained or interpreted due to the configuration of a number of different features, such as how people answer questions or which decision results they like. One way of explaining what each dimension means may be to label the dimension with these questions / answers and to find the question and answer that most correlates with the different extremes of the dimension.

In an embodiment, one way to cluster users may be to take the initial random grouping of users and to repeatedly move users between clusters to minimize how many users are different from each other in their own cluster. After several iterations, the process may be stopped or the process may continue until a threshold amount of error is reached.

In an embodiment, the present invention may facilitate registration by user similarity. Providing a username, email address, numerical user id, etc. provides a list of other users who are similar or dissimilar in some way. For example, providing a Facebook username provides a ranked list of other Facebook users with similar preference degrees in whole or in some specific manner as in electronic devices. In addition, this list may optionally be restricted to other users, for example, one degree apart from the first user in the social graph (e.g. by ranking my friends on Facebook by their similarity to me You can ask the person you are most like). In embodiments, user similarities may be computed via querying user questions about themselves to refer to their social graphs using contexts such as their location, IP address, time, and the like. Social graphs can be used by mapping users in social graphs to known users of different data sets based on heuristics for username, email address, first name, last name, birth date, address, gender and other similar information. Adjacent people can be searched within the social graph to consider more people even if they are removed more from the person you are trying to 'triangulate' on. For example, combining social graphs from Facebook with users who have reviews recorded on Amazon to find users on Facebook who are most similar to me, and then they tend to like on Amazon to give me laptop recommendations. Find a laptop. In another example, reviews may be filtered on sites like Yelp, TripAdvisor, Amazon, and the like based on people like you. This information may then be used to assist the user, such as "making friends" on Facebook and recommending the user to "follow" on Twitter. 30 and 31 provide examples of similarity profiles as may be provided by the present invention.

In an embodiment, the present invention may instead provide a level of dishonesty between recommendations by recommending people who like and / or dislike things. The present invention relates to things to purchase based on which similar users are brought to Amazon, locations to eat based on similar users on Yelp, Jagat, Foursquare, etc., and clicks on Google search results based on which similar people clicked. Etc. can be recommended. For example, consider the problem of recommending which users should follow new users to Twitter. The present invention can find users on Twitter and people following them and match some of these followers to the data set of the present invention to recognize things for each user based on the questions they answer about themselves. Can be. It can also match followers to other users, such as Amazon users, Yelp users, and so on, to learn other things about them. Based on this, the present invention may infer a follower of a particular user, such as a Twitter user. Now, new users can inquire about themselves and the followers of any existing user can find out which new user is most similar. The recommendation can then be generated for the new Twitter user to follow an existing Twitter user whose followers are most similar to the new user.

In an embodiment, the present invention immediately reflects new information from the user and, for example, makes real-time recommendations such as making immediate use of their context such as their social graphs, new facts about them, location changes, etc. Personalization can be facilitated to rerank recommendations or improve results in other ways.

In embodiments, the present invention may include location, time, weather, social graphs, and the like, such as using the location implicitly to indicate nearby places, drinks, sites they want to see, things to do, etc. that the user likes to eat. Can match based on a context such as For example, this process can then be used in a mobile application having access to location data via GPS. Optionally, recommendations may include current weather (for example, people don't recommend places they like only because of the courtyard when it rains), current time (for example, not recommending a nightclub at 10 am), calendar ( For example, users may be alerted by other parts of the user's context, such as busy and aware of where they should be in the future, social graphs (for example, a user's friend recommends a place on the same other). Can be.

In an embodiment, the present invention may provide a natural language question and answer interface, for example, to allow free-form or structured input from a user for a decision or recommendation they wish to assist. Input asks the user, allows other users to review the question, and uses automated techniques such as natural language processing (ie, "Is this question about electronic devices, travel, cars, or some other topic?" ), And the like.

In an embodiment, the present invention may find a user that is similar to a requesting user having expertise within a category of questions. For example, a user may have a similar degree of preference in a hotel and have a L.A. L.A. from people who perceive something about the hotel (such as self-described knowledge or proven knowledge based on their actions). You may want to get advice about hotels in your area. For example, the present invention may then alert these similar users to new questions and the need to help resolve them. Considerations may be given as to how many previous queries / alarms they have sent, how many have they already answered, how helpful their responses have been, and so on. Similar users may be allowed to engage in a conversation with the requesting user to help inform the requesting user of the decision or recommendation problem. The final conversation can be stored for others for use and encourage similar users to index the conversation in a structured form to assist subsequent recall by other requesting users.

The present invention can provide third party sites with the ability to learn about their users to find similar users and make recommendations independently, for example, in direct user interaction with the system. In embodiments, the present invention provides a preference and preference API that third parties, such as those hosting their own websites, may use to learn about user preferences, preferences, likes, dislikes, and other attributes. Where the user does not engage in conversation or interact directly with the computing equipment of the present invention. For example, a user may go to a website, such as Amazon.com, to query a product. In this case, Amazon enables the creation or enhancement of the user's preference and preference profile to better determine the user's preferences, preferences, and the like, thus allowing a third party to again send the user a meaningful response to the user query. It may have an API of the present invention that makes it possible to target well. In an embodiment, the third party may use the API to determine the user's preferences, preferences, etc. without user interaction, for example by determining their preferences and preferences through previous interactions with the user, where these transfers The interactions of in this example are from previous interactions while on a third party site such as Amazon, or from previous interactions with other websites hosting similar APIs or with direct interaction with the equipment of the present invention. Can be achieved through action. In an embodiment, an API may be used to identify similar users whose interaction with the user is different and thus may be used by multiple users to use selection, decision, election, recommendation, and the like. In embodiments, these other similar users may be associated with third party hosting APIs, or from other host API sites or from the equipment of the present invention. In embodiments, the use of the API of the present invention hosted by a third party site may provide a significant advantage to the third party site for recommendations related to the user's preferences, preferences, likes, dislikes, attributes, etc. .

In embodiments, the user's preferences and preferences may be determined or augmented by other similar users, other users connected to users in social networks, other users, such as other users, friends or family members who are related through personal or professional activities, and the like. Can be. In an embodiment, this can be done without having to query the user's question. For example, users may have existing preferences and preferences as established through the present invention, and this profile may be enhanced by gathering or inferring information about other users, such as their social networks, family, places of business, and the like. Can be. In an embodiment, the user's profile may be enhanced by inferring additional preferences and preferences from other similar users or users who have been shown to have some connection to the user, such as through a social network. In embodiments, the user's profile may be enhanced through selections made by other similar users, such as product selections, recommendations, and the like. In an embodiment, the system may determine the preferences of a known user through the use of natural language processing, such as inferring preference profiles by analyzing their rating of those they like and dislike or by analyzing how the user tags their user profile. Learn about profiles.

Referring to FIG. 32, in an embodiment, the present invention, when executed on one or more computers, may (1) provide a user preference learning API to a third party website 3220 to provide a user 3218 as applied to the market of third parties. Determining preferences of the user, wherein the preference learning API is executed as an extension of the computer equipment (3204), (2) receiving third party information related to the market of the third party (3208), (3) user Collecting (3218) preferences and storing them as user preference profiles (3210), (4) receiving a query from a user at a third party website 3220 associated with a third party's marketplace, and (5) To provide a recommendation to a third party based on the user preference profile and third party information to assist the third party in answering the received query to learn about the user through the use of computer equipment 3202. It is possible to provide a computer program product embodied on a computer readable medium that helps third party websites. In an embodiment, preference determination may be through the use of natural language processing. The computing equipment may be machine learning equipment. The third party information may consist of at least one of product information from a product manufacturer, product information from a web merchant, price information from another website, availability information from another website, reviews, comments, and ratings. The preference learning API may enable the collection of at least one of cost information, product information, personal information, and topic information. In addition, user profile preferences may be based on information inferred from the user's social network, where the user may receive additional conversations between the user and computer equipment.

Referring to FIG. 33, the present invention provides for the use of preferences and preferences to target responses returned to users to target advertisements, display reviews from similar users, recommend products or services, and You can display similar people on your network, and rank search results based on which results were most clicked by similar users. In an embodiment, the invention, when executed on one or more computers, comprises: (1) providing a user preference learning API to a third party website 3320 to determine a user's preference as related to the market of the third party, Wherein the preference learning API is executed as an extension of the computer equipment (3304), (2) receiving third party information related to the market of the third party (3308), (3) collecting the preferences of the user 3318 and Storing them as a user preference profile (3310), (4) receiving a query from the user (3318) at a third party website (3312), and (5) third party information related to the market of the third party and Using the user's 3318 preferences as stored in the user preference profile in a user preference learning API to perform a step of providing the user with a response related to the query from the user again. Through the use of non-3302 may provide a computer program product embodied in a computer readable medium that help to targeted the response returned to the user. In an embodiment, the computing equipment may be machine learning equipment.

The response may be to provide an advertisement to the user, where the advertisement may be based on the user's preference as stored in the user preference profile. Advertisements are provided by computer equipment, provided through third parties, enabled through preferences provided by third parties from computer equipment, delivered to other users in a user's social network, and the like. The response may provide a recommendation of a product, service, etc. related to the market of the third party. Collecting at least the preferences of the second user may form a user preference profile for the second user, determining the second user to be similar to the user based on the comparison of the preference profiles. The response may be to provide a recommendation made by the second user. The collection of preferences for the second user may be taken from an internet based social configuration, and the response may provide information to the user indicating the second user as a similar person to the configuration, where the internet based social configuration may be a social network. . The collection of preferences for at least the second user may include a search result selection, the query may be a search result, and the response may be a search result ranked according to at least the second user's search result selection. Collection may be from the selection of recommendations, purchases, and search results made by the user. The collection may be from a source that reveals the user's location behavior. The source may be user location information such as from web services Foursquare, Yelp, Google, Gowala, Facebook, and the like. The source may be user location information from the service provider. Third party information may include product manufacturers, product information from web merchants, price information from other websites, availability information from other websites, price information from merchants, availability information from merchants, reviews, comments and It may be made of at least one of the ratings. Preference determination may be through the use of natural language processing. The API may enable the collection of cost information, product information, personal information, topic information, and the like. The collection may come from user interaction as represented in an internet based social interaction construct, where the internet based social interaction construct may be a social network. The response may be a list of reviews categorized by the similarity of the review author to the user reading the review. Collecting user preferences may be through a third party website. Collection may be via crawling third party websites.

In an embodiment, the present invention may utilize a preference and preference API that a third party can use to target an advertisement to a user based on the user's preferences, where the user engages in a conversation or interacts directly with the computing equipment of the present invention. It doesn't work. For example, a user may have previously experienced a conversation of questions and answers with the system, through which the system can develop preference and preference profiles for the user. In an embodiment, the conversation may be provided directly by the equipment of the present invention or through a third party API provided by the present invention. Alternatively, the user may never interact with the equipment of the present invention, where the user's preferences and preference profiles may be created and updated through the user's interactions, responses, recommendations, reviews, and the like. In an embodiment, the system may determine the user's knowledge of a known user through their rating of things they like and dislike, or through the use of natural language processing, such as inferring a preference profile by analyzing how the user tags their user profile. You can also learn about preference profiles. The preference and preference profile for the user may then be used to target the advertisement to the user, such as an advertisement that matches the user's preference and preference. For example, the third-party preference and preference APIs are L.L. It may be associated with outdoor store websites such as L. L. Bean, REI, EMS, etc., where the outdoor store attempts to improve their targeting of advertisements to their customers. The customer can then visit the outdoor store website and make a query about a product such as hiking boots. The preference and preference APIs may then enable investigation of the user's preferences and preferences in order to establish a match for the placement of the ad in the user's browser. In this example, the user's preference and preference profile may indicate that the user enjoys traveling to New England, likes camping, has a family with children, and the like. As a result, the advertising equipment associated with the website may select an advertisement using information from the user's query, in this case hiking boots and information from their preference and preference profile. In this case the advertisement may be for a stay in White Mountain, which combines their preferences and the user's hiking boots query for a trip in New England. In addition, accommodation can be family accommodation because the user likes to travel as a family and has attributes that reflect the user's preferences. In embodiments, third party preference and preference APIs may enable third parties to enhance their targeting of ads to users so that they may be able to increase revenues made for a given ad placement from an advertising sponsor. . In embodiments, user preferences and preference profiles may be developed in real time as focusing on their immediate queries. Preferences and preferences may be collected to better target advertisements to the user, such as during subsequent improvements in product search at points of purchase and the like.

In an embodiment, the advertisement may be targeted to a group of individuals associated with the user based on preferences and preferences inferred through the user or the user's social network. For example, the preference and preference APIs used by third parties can be used to set preferences and preferences for groups, node clusters, etc., such as in a user's social network. In embodiments, preferences and preferences inferred from social networks may utilize preferences and preference profiles previously formed through third party sites or through equipment directly related to the present invention. These preferences and preferences can then be used to better target the advertisement to the user or a member of the user's social network. In an example, a third party may want to target an advertisement to a user, where the user has a set preference and preference profile stored in the equipment of the present invention. The third party may then use the information in the user's profile to target the advertisement. Alternatively, the third party may additionally use information inferred from the social network in which the user is part, such as information pertaining to the topic of the social network, the user's common interest in relation to the user in the social network, and the like. For example, a user may have a preference and preference profile indicating that they are middle aged, politically conservative, in the countryside, etc., and are associated with the user in a social network with a predominant interest. In this example, the third party may target an advertisement for shooting equipment, shooting trips, etc., wherein the advertisement is selected based on the user's existing preference and preference profile in addition to inference from the user's social network. In an embodiment, this does not need to be coupled to the user in a conversation as described herein, but rather directly through the user's interaction at a third party website, where the present invention has a third party API or the like. Can be performed through third party supplied information. In an embodiment, the third party may also use preferences and preferences from the user to target the advertisement to other members of the user's social network.

In an embodiment, advertising to the user may be targeted based on the user's product selection, recommendations, etc., having similar preferences and preferences. For example, the first user may have similar preferences and preferences to the second user, where the first user has an existing preference and preference profile and makes specific product selections, recommendations, and the like. The advertisement may then be targeted to the second user based on the determination of the first user. For example, a first user may have a profile indicating that they are old, retired, live in California, enjoy travel, etc., who have previously made product selection for baggage. The second user may then be offered a recommendation for similar baggage based on the similarity of the preferences and preferences of the two users. In an embodiment, this can be done without the need to join one of the users in the conversation.

In an embodiment, the invention, when executed on one or more computers, comprises: (1) providing a user preference learning API to a third party website to determine a user's preferences as applied to a third party's marketplace. The API is executed as an extension of computer equipment, (2) receiving third party information related to the market of the third party, (3) collecting user preferences and storing them as a user preference profile, (4) Receiving a query from the user at a third party website associated with the market of the third party, and (5) providing an advertisement to the user, wherein the advertisement is based on the identified preferences of the user The use can provide a computer program product embodied in a computer readable medium that helps target an advertisement to a user. In an embodiment, determining the preference may be through the use of natural language processing. Advertisements may be provided by computer equipment. Advertisements may be provided through third parties and may be enabled through preferences provided to third parties from computer equipment. The API may enable the collection of at least one of cost information, product information, personal information and topic information. The decision may also be based on information inferred from the user's social network. Advertisements may be delivered to other users associated with the user, such as through social networks.

In an embodiment, the present invention may provide a preference and preference API that a third party can use to provide a review from a similar user to the user, where the user and the similar user engage in a conversation or interact directly with the computing equipment of the present invention. It may not work. For example, the preference and preference API allows the present invention to collect preference and preference information for a user, to provide third party with preference and preference information for the user from a previously set preference and preference profile, and the user's short-term behavior. May provide the third party with preference and preference information for the user. In an embodiment, the user may never interact with the equipment of the present invention, where the user's preferences and preference profiles may be created and updated through the user's interactions, responses, recommendations, reviews, and the like. In an embodiment, the system may be based on the use of a known user through the use of natural language processing, such as inferring a preference profile through their ratings on things they like and dislike, or by analyzing how users tag their user profiles. You can also learn about preference profiles. In this case, a similar user may have a preference and preference profile set previously, and thus have a preference and preference profile that can be matched to the user. In addition, these similar users may have reviews associated with their profiles. The system can now match the user to similar users, and then provide the user with relevant reviews. For example, a user may have an existing preference and preference profile by the system, such as, for example, directly to computer equipment or through at least one third party API, and how other similar users may have a given product, service, person You may want to know if you consider events, events, and so on. The system can then retrieve the preference and preference profiles for similar users for the subject that the current user is interested in. In this way, the system may now be able to provide reviews, etc., to the current user from similar users, thus helping to determine what the current user may want to do based on their preference and preference. For example, a user may proceed to a product website utilizing the preference and preference API of the present invention and is interested in a review of a digital camera. The third party can now find similar users, and then can retrieve digital camera reviews by those similar users and provide the review to the current user. In an embodiment, the review may be present in third party equipment, in another third party equipment, in the equipment of the present invention, or the like. In an embodiment, the ability to display reviews of similar users allows the user to access more relevant reviews in a more time efficient manner, and third party users of the API provide more targeted and relevant support to their users. It may be possible to.

In an embodiment, the present invention may provide a preference and preference API that allows a third party to provide a user with a review from a similar user, where the user is determined to be similar without a similar user participating in a conversation through the present invention. . For example, similar users may be identified as similar through social networks, friends, family, work, and the like. In an example, a user may be associated with a second user through a social network determined to be 'similar', such as through age, interest, and the like. Similar users may then provide reviews for products, activities, etc., for example. This review may then be presented to the user as related through similarity to other users. In embodiments, similar users may be determined through similar recommendations on other topics, for example in combination with other factors, where the other factors may be socially relevant.

In an embodiment, the invention, when executed on one or more computers, comprises: (1) providing a user preference learning API to a third party website to determine a user's preferences, wherein the preference learning API is implemented as an extension of computer equipment. (2) collecting the preferences of the plurality of users, wherein the plurality of users comprises the user, and (3) storing the user's preferences in the preference and preferences database comprising the plurality of preferences and preference profiles. (4) receiving a request from the user via a third party preference and preference learning API for topic review from a user with similar preference and preferences, (5) a preference of at least one other user in the preference and preference database Matching the user's preferences to (6) reviewing topics from other matched users. A computer program embodied in a computer readable medium that assists a user in discovering similar user reviews through the use of computer equipment by performing the steps of retrieving reviews related to the request, and (7) providing reviews to the user. Products can be provided. In an embodiment, determining the preference may be through the use of natural language processing. Reviews can be found within computer equipment, third party equipment, and the like. The computing equipment may be machine learning equipment. The preference learning API may enable the collection of cost information, product information, personal information, topic information, and the like. The review may be provided by a similar user who does not have a preference and preference profile, where the user may be similar to that determined by social relevance where social relevance may be a social network.

In embodiments, the present invention may provide preference and preference APIs that allow third parties to use recommended products, services, and the like. For example, a user may come to a third party website upon retrieval of a recommendation for a product, and the third party may then use the preference and preference APIs to better understand what the user typically likes From this preference, the product can be proposed. In an example, a user may come to an audio store website to find a recommendation for an audio system to play music from their iPhone. The third party may then use existing preferences and preferences for the user via the API. In this example, the user's preferences and preferences may indicate that they are college students and their social life is busy. From this information, the third party can now make a recommendation, such as a recommendation for an audio system that is portable, small, powerful, and the like. Alternatively, third parties can use the preference and preference APIs to determine their preferences and preferences in user queries, such as those generally targeted to users, content of third parties, or the like. Third parties can now make recommendations using this new preference and preference information alone or in combination with previous preference and preference profiles through the present invention. The preferences and preferences as set through third parties can now be stored within the equipment of the present invention, for example in combination with new preferences and preference profiles generated via other third party APIs or directly through the equipment of the present invention. Or may be used again. In an embodiment, the ability to use preference and preference APIs may improve recommendations for products, services, etc. made through third party sites.

In embodiments, the present invention may provide preference and preference APIs that third parties can use to recommend products, services, and the like to users based on similar user actions. For example, two users may have a previously set preference and preference profile of the present invention, where one of the users selects a product, a service, and the like, and a third party is now those determined as determined through their profile. The recommendation can be provided to other users based on the similarity of. In an example, two users may be determined to be similar through their preference and preference profile, for example by their age, location, political perspective, social activity, and the like. The first user can then select a product, such as a car. In the case where a second similar user is required to direct attention to a vehicle, such as through search, advertisement selection, or explicit questioning in the present invention, the present invention provides a vehicle selection to the second user as a potential suitability due to their similarity. Can be provided. In an embodiment, this may be done without a conversation provided to one or both users.

In an embodiment, the invention, when executed on one or more computers, comprises: (1) providing a user preference learning API to a third party website to determine a user's preferences as applied to the third party's products and services, The preference learning API is implemented as an extension of computer equipment, (2) receiving third party information related to third party products and services, and (3) collecting user preferences and storing them as user preference profiles. (4) receiving a query from the user at a third party website associated with at least one of the third party's products and services, and (5) providing a recommendation for the at least one of the products and services to the user from the computing equipment; As a step, the recommendation is performed by the user through the use of computer equipment by performing a step based on the query and the user's identified preferences. It can provide a computer program product embodied on a computer readable medium that helps to find the others. In an embodiment, the determination of preference may be through the use of natural language processing. The collection may be from at least one of a third party website on the internet, a plurality of third party websites on the internet, a recommendation made by a user, selection of purchases and search results, and the like. The computing equipment may be machine learning equipment. Third party information may include product information from product manufacturers, product information from web merchants, price information from other websites, availability information from other websites, price information from merchants, availability information from merchants, Reviews, comments, ratings and the like. The API may enable the collection of cost information, product information, personal information, topic information, and the like. Preference may be derived from the motion of a second similar user, where similarity may be determined through preference and preference profiles for the user and the second similar user. The action of the second similar user may be the selection of at least one of a product and a service.

In embodiments, the present invention may provide preference and preference APIs that third-party social network sites may use to represent people who are similar to them on a social network. These similar people can be displayed as a list, by photo, by region, age, gender, and the like. For example, a user may come to a social network site and may be asked to view or access people similar to them. Social network sites may then use preference and preference APIs to provide the user with a conversation to determine their preferences and preferences, such as social status, social networking, activity, music, personality, and the like. Alternatively, the user may have in advance a preference and preference profile as determined directly by the equipment of the present invention via other third party APIs, via a social network, and so on. The social network is then used to match users to others on the social network, such as through other people's preferences and preference profiles, as determined previously, through information available about others as available through the social network, and so forth. You can use this information. For example, the user's preferences and preferences can indicate that they are young and enjoy going to a club in NYC. Social networking sites may now be able to match users to similar people on social networks, such as by list, by photo, by category, by region of a city, and the like. In embodiments, preference and preference APIs with a social network may provide an improved matching experience for users who attempt to discover other similarly socially related people.

In an embodiment, the invention, when executed on one or more computers, comprises: (1) receiving an initial request from a user via a third party social networking site API, where the initial request is to discover other similar users to them in the social network. (2) identifying a user's preferences via a social networking site API, (3) matching user preferences to other users on a social network having users with similar preferences, and (4) A computer embodied in a computer readable medium that assists a user in discovering other similar users on a social networking site through the use of computer equipment by performing a step of providing a match request to a user, including another user that matches the user's preferences. Program products can be provided. In an embodiment, confirmation of preference may be through the use of natural language processing. The matching result can be presented to the user as a list of similar users. The matching result can be presented to the user as a profile of similar users. The matching result can be presented to the user as a link to a similar user in the social network. The computing equipment may be machine learning equipment.

In an embodiment, the present invention may provide a preference and preference API that third-party search equipment may use to rank search results based on which results were most selected by similar users. For example, the search equipment may provide the user with an opportunity to improve the relevance of how search results are listed via the preference and preference profile, as provided through the preference and preference API of the present invention. The preference and preference profile database, etc. may then be accumulated and maintained, from which the search equipment may rank the search results for the user with the results previously selected by other similar users. In an example, a user may have a preference and preference profile that represents a retired male who likes to navigate and is slightly adventurous. When a user searches for Caribbean vacation destinations, the search equipment may rank search results with these preferences and preference attributes listed first, such as, for example, sailing boat rental packages on the island, hiking on the island, exotic destinations, and the like. In an embodiment, the use of preference and preference APIs provided to the search equipment may improve the relevance of the ranked search results to the user.

In an embodiment, the invention, when executed on one or more computers, comprises (1) receiving a search request from a user via a third party search device, and (2) confirming a user's preferences, wherein the preferences identified by the user are Creates a preference and preference profile for a user and is stored in a preference and preference storage device comprising a plurality of different user preference and preference profiles, where the profile also includes a history of search results selected by other users in a previous search (3) matching the user to another user with similar preference and preference profile, (4) determining a search result set for the user's search request, (5) another user with similar preference and preference profile Matching the search result set to a history of the search results selected by (6) Embodied in a computer readable medium that assists in ranking the search results through the use of computer equipment by performing the steps of providing the user with search results ranked according to the matched results selected by other users having similar preference and preference profiles. A computer program product can be provided. In an embodiment, confirmation of preference may be through the use of natural language processing. The computing equipment may be machine learning equipment. The search equipment may be a search engine.

Referring to FIG. 34, the present invention may use a social graph to infer preference and preference for an unknown user by finding a path through an Internet-based social interaction configuration to a person having a known preference. In this way, the present invention may provide a way for the system to obtain data for a user who has not previously heard. In an embodiment, the invention, when executed on one or more computers, (1) identifying preferences of a plurality of users that are part of an Internet-based social interaction configuration, wherein the plurality of users become a plurality of known users ( 3404) (2) determining an internet based social interaction graphical representation 3412 for the plurality of known users, and (3) based on the correlation between the plurality of known users and the unknown user in the graphical representation. Deriving the preferences of unknown users present in the Internet-based social interaction graphical representation 3412 of a plurality of known users may be performed to perform the use of the Internet-based social interactive graphical representation on the computer equipment 3402. Provide a computer program product embodied in a computer readable medium that helps determine preferences of unknown users. There. In an embodiment, the internet-based social interaction graphical representation may be a social network, a social graph, a social diagram, or the like. The unknown user may be spaced 3 degrees, 5 degrees, etc. from the nearest known user in the internet-based social interaction graphical representation. The inferred preference of the unknown user can cause the unknown user to become a new known user, and the new known user can be used to contribute to the inference of the preference of the second unknown user. Preference may include personal information, topic information, etc. related to the user's interaction, where the interaction may be via an internet-based social interaction graphical representation. Interaction may be through an API provided on a third party website. Inference can be provided with other known users associated with the user in the user's Internet-based social interaction configuration. Confirmation of preference may be through the use of natural language processing. The computing equipment may be machine learning equipment. Inferred preferences can be used to target ads to unknown users in order to share reviews with unknown users. Inferred preferences can be used to recommend products, services, etc. to unknown users. Inferred preferences can be used to assist in ranking search results for unknown users. Known users in close proximity to the unknown user may pass more weight to the inference algorithm. Inferred preferences can be improved by information from other sources, where the other sources are third party sources, recommendations made by multiple known users, search queries by multiple known users, and among multiple known users. Selection of one search result, personal preference as determined through web interaction by at least one of the plurality of known users, and the like. Other sources may include third party preference learning APIs.

In embodiments, the present invention may infer preferences and preferences for unknown users by using social network graphs, diagrams, graphical representations, and the like to find a path to people with known preferences through social networks, Or vice versa. Social diagrams are generally speaking, mapping of multiple users and how they are related. By using social diagrams, preferences and preferences of known and unknown users can be determined from their correlation in the diagram. For example, a user with a known preference and preference profile may be directly associated with a plurality of other users as represented in the social diagram. For the first approximation, it can be assumed that these plurality of other users are similar to the user and thus have similar preferences and preferences. These other users may then provide an improved service that takes advantage of recognizing the user's preferences and preferences as described herein. For example, if a user is provided with a known preference and preference profile indicating that they are rock climbers, the user in the first link of the user's social diagram may also be assumed to be a rock climber. In practice, this may prove to be a too general assumption. However, it may be a good assumption that a user has relevance to other rock climbers, so the system can proceed through social diagrams searching for other known users who enjoy rock climbing. In this example, it may be found that there are other known users, such as three link-spaced ones, who also enjoy rock climbing, which are found in a cluster that connects to the first user. From this, it may be a good assumption that this cluster is a group of rock climbers and that rock climbers can all share a set of similar preferences and preferences. In embodiments, preference and preference may be inferred from relevance in social network diagrams, and thus may benefit from the present invention as described herein.

In an embodiment, the present invention, when executed on one or more computers, includes (1) identifying a user's preferences, wherein the user becomes a known user, and (2) determines a social network graphical representation of the known user. (3) determining presence in other known users in the social network graphical representation of the known user, and (4) based on the correlation between the unknown user and the known users and other known users in the network graphical representation. Deducing the preferences of the unknown user present in the social network graphical representation of the known user, thereby helping to determine the preference and preference of the unknown user through the use of the social network graphical representation on the computer equipment. A computer program product embodied in the medium may be provided. In an embodiment, confirmation of preference may be through the use of natural language processing. The social network graphical representation may be a social graph, a social diagram, or the like. The computing equipment may be machine learning equipment.

In an embodiment, the present invention may enhance the recommendations provided through two or more third party APIs by combining the user's preferences and preferences as determined through two or more third party APIs. For example, there may be a preference and preference profile set via more than one third party API, and by combining these different preference and preference profiles, the combined preference and preference profile may be generated by the present invention. In addition, as additional preference and preference profiles are created via third party APIs, they can be used to continuously update the combined preference and preference profile for the user. The third party can then use the combined preference and preference profile to improve their recommendations. This may be especially true when different third parties focus on preference and preference profiling in different areas such as products, personal relationships, services, entertainers, and the like. It can be appreciated that combining multiple more specific profiles into a combined profile can provide a richer preference and preference profile, which can then be generated via any one of the more specific profiles. In addition, users can change their preferences and preferences over time, so combining more recent user profile interactions on one third party API means that the user has not interacted recently but It may be beneficial to other third parties if they want to keep their user profile up to date.

Referring to FIG. 35, in an embodiment, the present invention, when executed on one or more computers, may (1) communicate with a user via a first third party website 3514 via a first preference learning third party API. Generating 3504 a first preference and preference profile of the user 3512, (2) collecting additional user interaction information via a second third party website 3514 via a second third party API 3508, and (3) combining the additional user interaction information with the preference and preference profile to enhance the preference and preference profile 3510 to thereby enhance the preference and preference profiling through the use of computer equipment. A computer program product embodied in a computer readable medium for providing 3502. In an embodiment, generating the first preference and preference profile may be through confirmation of user preference through the use of natural language processing. The computing equipment may be machine learning equipment. The API may enable the collection of at least one of cost information, product information, personal information and topic information.

The methods and systems described herein may be partially or wholly deployed through a machine executing computer software, program code and / or instructions on a processor. The invention may be embodied as a method on a machine, as part of a machine or as a system or apparatus associated with a machine, or as a computer program product embodied on a computer readable medium executing on one or more of the machines. The processor may be part of a server, client, network infrastructure, mobile computing platform, stationary computing platform, or other computing platform. A processor may be any kind of computing or processing device capable of executing program instructions, code, binary instructions, and the like. A processor may be a signal processor, digital processor, embedded processor, microprocessor, or coprocessor (math coprocessor, graphics coprocessor, communication coprocessor, etc.) that may directly or indirectly facilitate execution of program code or program instructions stored thereon. May be any variation, such as or the like. In addition, the processor may enable execution of multiple programs, threads, and code. Threads can run concurrently to improve processor performance and to facilitate concurrent operation of applications. As an implementation, the methods, program code, program instructions, and the like described herein may be implemented in one or more threads. Threads may spawn other threads that may have assigned priorities associated with them, and the processor may execute these threads based on priority or any other order based on instructions provided within the program code. The processor may include a memory that stores methods, codes, instructions, and programs as described herein and elsewhere. The processor may access the storage medium through an interface capable of storing methods, codes, and instructions as described herein and elsewhere. Storage media associated with the processor for storing methods, programs, code, program instructions, or other types of instructions that may be executed by a computing or processing device include, but are not limited to, CD-ROMs, DVDs, memories, hard disks. , Flash drive, RAM, ROM, cache, and the like.

A processor may include one or more cores capable of improving the speed and performance of the multiprocessor. In an embodiment, the process may be a dual core processor, quad core processor, another chip-level multiprocessor, or the like, combining two or more independent cores (called dies).

The methods and systems described herein may be deployed in part or in whole through a server, client, firewall, gateway, hub, router, or other machine running computer software on such computer and / or networking hardware. Software programs may be associated with servers that may include file servers, print servers, Internet servers, intranet servers, and other variations such as secondary servers, host servers, distributed servers, and the like. The server includes one or more of the interfaces capable of accessing other servers, clients, machines, and devices through memory, processors, computer readable media, storage media, ports (physical and virtual), communication devices, and wired or wireless media, and the like. can do. The methods, programs or codes as described herein and elsewhere may be executed by the server. In addition, other devices required for the execution of the method as described in this application may be considered as part of the infrastructure associated with the server.

The server may provide an interface to other devices including, but not limited to, clients, other servers, printers, database servers, print servers, file servers, communication servers, distributed servers, and the like. In addition, this coupling and / or connection may facilitate remote execution of the program across the network. Networking of some or all of these devices may facilitate parallel processing of a program or method at one or more locations without departing from the scope of the present invention. In addition, any device connected to the server via an interface may include at least one storage medium capable of storing methods, programs, code, and / or instructions. The central repository may provide program instructions to be executed on different devices. In this implementation, the remote storage can act as a storage medium for program code, instructions, and programs.

The software program may be associated with a client, which may include file clients, print clients, domain clients, Internet clients, intranet clients and other variants such as secondary clients, host clients, distributed clients, and the like. Clients include one or more of memory, processors, computer readable media, storage media, ports (physical and virtual), communication devices, and interfaces capable of accessing other clients, servers, machines, and devices through wired or wireless media, and the like. can do. The methods, programs or codes as described herein and elsewhere may be executed by the client. In addition, other devices required for the execution of the method as described in this application may be considered as part of the infrastructure associated with the client.

Clients may provide interfaces to other devices including, but not limited to, servers, other clients, printers, database servers, print servers, file servers, communication servers, distributed servers, and the like. In addition, this coupling and / or connection may facilitate remote execution of the program across the network. Networking of some or all of these devices may facilitate parallel processing of a program or method at one or more locations without departing from the scope of the present invention. In addition, any device connected to the client via an interface may include at least one storage medium capable of storing methods, programs, applications, code, and / or instructions. The central repository may provide program instructions to be executed on different devices. In this implementation, the remote storage can act as a storage medium for program code, instructions, and programs.

The methods and systems described herein may be partially or wholly deployed throughout the network infrastructure. The network infrastructure may include elements such as computing devices, servers, routers, hubs, firewalls, clients, personal computers, communication devices, routing devices, and other active and passive devices, modules, and / or components as are known in the art. It may include. The computing and / or non-computing device (s) associated with the network infrastructure may include storage media such as flash memory, buffers, stacks, RAM, ROM, etc., separately from other components. The processes, methods, program code, instructions described herein and elsewhere may be executed by one or more of the network infrastructure elements.

The methods, program codes, and instructions described herein and elsewhere may be implemented on a cellular network having multiple cells. The cellular network may be a frequency division multiple access (FDMA) network or a code division multiple access (CDMA) network. Cellular networks may include mobile devices, cell sites, base stations, repeaters, antennas, towers, and the like. The cell network may be GSM, GPRS, 3G, EVDO, mesh or other network type.

The methods, program codes, and instructions described herein and elsewhere may be implemented on or through a mobile device. Mobile devices may include navigation devices, cell phones, mobile phones, mobile personal digital assistants, laptops, palmtops, netbooks, pagers, electronic book readers, music players, and the like. These devices, apart from other components, may include storage media such as flash memory, buffers, RAM, ROM, and one or more computing devices. Computing devices associated with the mobile device may be enabled to execute program code, methods, and instructions stored thereon. Alternatively, the mobile device can be configured to execute the command in cooperation with another device. The mobile device can communicate with a base station interfaced with a server and configured to execute program code. The mobile device can communicate on a peer-to-peer network, a mesh network, or another communication network. The program code may be stored on a storage medium associated with the server and executed by a computing device embedded in the server. The base station can include a computing device and a storage medium. The storage device can store program code and instructions executed by the computing device associated with the base station.

Computer software, program code and / or instructions may be computer components, devices, and recording media that hold digital data used for computing for a predetermined time interval, semiconductor storage known as random access memory (RAM), and typically optical Mass storage for more permanent storage devices such as disks, hard disks, tapes, drums, cards and other types of magnetic storage devices, processor registers, cache memory, volatile memory, nonvolatile memory, CD, Optical storage devices such as DVDs, and removals such as flash memory (e.g. USB sticks or keys), floppy disks, magnetic tapes, paper tapes, punch cards, standalone RAM disks, Zip drives, removable mass storage devices, offline Media, dynamic memory, static memory, read / write storage, variable storage, reading Storage on machine readable media that can include other computer memories such as memory, random access, sequential access, location addressable, file addressable, content addressable, network attached storage, storage area network, bar code, magnetic ink, and the like. And / or can be accessed.

The methods and systems described herein can transform physical and / or intangible items from one state to another. The methods and systems described herein can also transform data representing physical and / or intangible items from one state to another.

The elements described and illustrated herein, including flow diagrams and block diagrams throughout the figures, imply logical boundaries between the elements. However, according to a software or hardware engineering implementation, the illustrated elements and their functions may be programs stored thereon as monolithic software structures, as standalone software modules, or as modules using external routines, code, services, etc., or any combination thereof. It may be implemented on a machine via a computer executable medium having a processor capable of executing instructions, all such embodiments may be within the scope of the present invention. Examples of such machines include, but are not limited to, personal digital assistants, laptops, personal computers, mobile phones, other portable computing devices, medical equipment, wired or wireless communication devices, transducers, chips, calculators, satellites, tablet PCs, Electronic books, gadgets, electronic devices, artificial intelligence devices, computing devices, networking equipment, servers, routers, and the like. Moreover, the elements shown in the flow diagrams and block diagrams or any other logical component may be implemented on a machine capable of executing program instructions. Thus, while the drawings and descriptions describe functional aspects of the disclosed systems, no specific arrangement of software for implementing these functional aspects should be inferred from these descriptions unless explicitly stated or otherwise apparent from the context. Similarly, it will be understood that the various steps identified and described above may be changed and that the order of steps may be applied to specific applications of the techniques disclosed herein. All such variations and modifications are intended to be within the scope of this specification. As such, the illustration and / or description of the order for the various steps is understood to require a specific order of execution for these steps unless otherwise required by a particular application, not explicitly stated, or otherwise apparent from the context. It should not be.

The above described methods and / or processes and their steps may be realized in hardware, software or any combination of hardware and software suitable for a particular application. The hardware may include a general purpose computer and / or a dedicated computing device or a specific aspect or component of a specific computing device or a specific computing device. The process may be realized in one or more microprocessors, microcontrollers, embedded microcontrollers, programmable digital signal processors or other programmable devices, along with internal and / or external memory. The process may also be embodied in or instead of application specific integrated circuits, programmable gate arrays, programmable array logic or any other device or combination of devices that may be configured to process electronic signals. One or more of the processes may be realized as computer executable code capable of being executed on a machine readable medium.

Computer-executable code may be stored, compiled or interpreted to execute on one of the devices as well as on a heterogeneous combination of processors, processor architecture or combination of different hardware and software or on any other machine capable of executing program instructions. Can be generated using a structured programming language such as C, an object-oriented programming language such as C ++, or any other high or low level programming language (assembly language, hardware description language and database programming language and technology).

Thus, in one aspect, each of the methods described above and combinations thereof may be embodied in computer executable code that performs their steps when executed on one or more computing devices. In other aspects, the methods may be embodied in a system that performs these steps, may be distributed across devices in a number of ways, or all functionality may be integrated into dedicated, standalone devices, or other hardware. In another aspect, the means for performing the steps associated with the process described above may include any hardware and / or software described above. All such substitutions and combinations are intended to be within the scope of the present invention.

Although the present invention has been disclosed in connection with the preferred embodiments shown and described in detail, various modifications and improvements will be readily apparent to those skilled in the art. Accordingly, the spirit and scope of the present invention should not be limited by the above examples, but should be understood as the widest concept acceptable by law.

All documents referred to herein are incorporated herein by reference.

Claims (20)

A machine comprising a memory and at least one processor,
Computer equipment executable by the machine,
The computer equipment,
Access social graph information that identifies one or more users connected to a particular user in the social graph,
Identify preferences of the one or more users,
And based on the identified preferences of the one or more users, generate a user preference profile indicative of the preferences of the particular user.
system.
The method according to claim 1,
The computer equipment also,
Determine that the particular user is interested in a particular topic,
Determine, via the social graph, a second user connected to the particular user is interested in the particular topic,
Identify a cluster of users connected to both the particular user and the second user through the social graph,
Infer that each user in the cluster of users is interested in the particular topic.
system.
3. The method of claim 2,
The second user is spaced at least a predetermined number of degrees from the particular user in the social graph
system.
3. The method of claim 2,
The second user is at least 3 degrees away from the particular user in the social graph
system.
The method according to claim 1,
The social graph is associated with online social network services
system. The method according to claim 1,
At least one of the one or more users is at least 3 degrees apart from the particular user in the social graph
system.
The method according to claim 1,
At least one of the one or more users is at least 5 degrees apart from the particular user in the social graph
system.
The method according to claim 1,
The preferences of the one or more users include at least one of topic information and personal information related to the interaction of the one or more users.
system.
The method according to claim 1,
Among the one or more users, a user close to the specific user carries more weight in the inference algorithm.
system.
The method according to claim 1,
Inferred preferences of the particular user are used to target an advertisement for the particular user.
system.
The method according to claim 1,
Inferred preferences of the particular user are used to share reviews with the particular user.
system.
The method according to claim 1,
The inferred preference of the particular user is used to recommend at least one of a product and a service to the particular user
system.
The method according to claim 1,
Inferred preferences of the particular user are used to assist in ranking the search results for that particular user.
system.
A non-transitory machine readable storage medium containing instructions, comprising:
The instructions, when executed by one or more processors of a machine, cause the machine to:
Accessing social graph information identifying one or more users connected to a particular user in the social graph,
Identifying preferences of the one or more users;
Based on the identified preferences of the one or more users, performing an operation comprising generating a user preference profile indicative of the preferences of the particular user.
A machine-readable storage medium.
15. The method of claim 14,
The operation includes:
Determining that the particular user is interested in a particular topic;
Determining, via the social graph, that a second user connected to the particular user is interested in the particular topic;
Identifying a cluster of users connected to both the particular user and the second user via the social graph;
Inferring that each user in the cluster of users is interested in the particular topic
A machine-readable storage medium.
The method of claim 15,
The second user is spaced at least a predetermined frequency from the particular user in the social graph
A machine-readable storage medium.
The method of claim 15,
The second user is at least 3 degrees away from the particular user in the social graph
A machine-readable storage medium.
15. A method implemented by a computer,
Accessing social graph information identifying one or more users connected to a particular user in the social graph,
Identifying preferences of the one or more users;
Based on the identified preferences of the one or more users, generating a user preference profile indicative of the preferences of the particular user.
A method implemented by a computer.
19. The method of claim 18,
Determining that the particular user is interested in a particular topic;
Determining, via the social graph, that a second user connected to the particular user is interested in the particular topic;
Identifying a cluster of users connected to both the particular user and the second user via the social graph;
Inferring that each user in the cluster of users is interested in the particular topic
A method implemented by a computer.
The method of claim 19,
The second user is spaced at least a predetermined frequency from the particular user in the social graph
A method implemented by a computer.

Images